STA301 — Final Term Summary (Lectures 23–45)

📘 Lecture 23 — Graphical Representation of the Distribution Function of a Discrete Random Variable

📖 Overview: This lecture explores the distribution function (cumulative distribution function) for discrete random variables, including its graphical representation as a step function. It then introduces the concept of mathematical expectation, covering the mean, variance, moments, and key properties of expected values — all essential for characterizing probability distributions.

🗂️ Topics Covered

The lecture begins with the definition and interpretation of the distribution function for discrete random variables, illustrated through the example of tossing three coins. It then presents another example involving defective clock radios. Following this, the concept of mathematical expectation is introduced, along with the expectation of functions of a random variable, variance, moments about the origin and about the mean, moment ratios for skewness and kurtosis, and finally, the properties of mathematical expectation.

📝 Lecture Summary

Graphical Representation of the Distribution Function of a Discrete Random Variable

The distribution function of a random variable X, denoted by F(x), is defined as F(x) = P(X ≤ x). It gives the probability that X takes a value less than or equal to a specified value x. It is also called the cumulative distribution function (cdf).

🔑 Definition — Distribution Function (F(x)): F(x) = P(X ≤ x), the cumulative probability of X from its smallest value up to x.

📌 Example: Find the probability distribution and distribution function for the number of heads when 3 balanced coins are tossed. The sample space S = {HHH, HHT, HTH, THH, HTT, THT, TTH, TTT}. Let X = number of heads. The probabilities are: f(0)=1/8, f(1)=3/8, f(2)=3/8, f(3)=1/8. The cumulative probabilities are: F(0)=1/8, F(1)=4/8, F(2)=7/8, F(3)=1. The distribution function is: F(x) = 0, for x < 0 F(x) = 1/8, for 0 ≤ x < 1 F(x) = 4/8, for 1 ≤ x < 2 F(x) = 7/8, for 2 ≤ x < 3 F(x) = 1, for x ≥ 3

The graph of the distribution function resembles a staircase and is known as a step function or jump function. For this example, the probability of at most one head is 4/8, and at most two heads is 7/8. 💡 Why this matters: The distribution function allows us to compute probabilities for ranges of values, not just individual points, which is fundamental for real-world decision-making.

📌 Example: A store places its last 15 clock radios in a clearance sale; 5 are defective. A customer tests 3 different radios selected at random. Let X = number of defective radios in the sample. The probability distribution is: P(X=0)=0.26, P(X=1)=0.49, P(X=2)=0.22, P(X=3)=0.02. The distribution function is: F(0)=0.26, F(1)=0.75, F(2)=0.97, F(3)≈1. The probability that the sample contains at most one defective radio is 0.75, and at most two defective radios is 0.97.

Mathematical Expectation

Let a discrete random variable X have possible values x₁, x₂, …, xₙ with corresponding probabilities f(x₁), f(x₂), …, f(xₙ) such that Σf(xᵢ)=1. The mathematical expectation or expected value of X, denoted by E(X), is defined as E(X) = x₁f(x₁) + x₂f(x₂) + … + xₙf(xₙ) = Σᵢ₌₁ⁿ xᵢ f(xᵢ). E(X) is also called the mean of X and is usually denoted by μ. It is a weighted mean of the variable’s possible values, each weighted by its probability. E(X) represents the average value of X over a very large number of trials.

🔑 Definition — Mathematical Expectation E(X): E(X) = Σ xᵢ f(xᵢ), the weighted average of all possible values of X, weighted by their probabilities.

📐 Formula: E(X) = Σᵢ xᵢ f(xᵢ) → The sum of each value of X multiplied by its probability.

📌 Example: An umbrella salesman earns $30 per day if it rains and loses $6 per day if it is fair. The probability of rain is 0.3. Let X = dollars earned. Possible values: 30 (rain, P=0.3) and -6 (no rain, P=0.7). E(X) = 30(0.3) + (-6)(0.7) = 9.0 - 4.2 = $4.80 per day. On average, the salesman expects to earn $4.80 per day.

Mean, Variance and Moments of a Discrete Probability Distribution

Let H(X) be a function of the random variable X. Then H(X) is also a random variable. The expected value of the function H(X) is E[H(X)] = Σ H(xᵢ) f(xᵢ). If H(X) = (X - μ)², then E(X - μ)² = Σ (xᵢ - μ)² f(xᵢ). This is called the variance, denoted by Var(X) or σ². The shortcut formula for variance is σ² = E(X²) - [E(X)]². The positive square root of the variance is the standard deviation.

If H(X) = Xᵏ, then E(Xᵏ) = Σ xᵢᵏ f(xᵢ) is called the kth moment about the origin, denoted by μ'ₖ. If H(X) = (X - μ)ᵏ, then μₖ = E(X - μ)ᵏ = Σ (xᵢ - μ)ᵏ f(xᵢ) is the kth moment about the mean.

Skewness is measured by β₁ = μ₃² / μ₂³, and kurtosis by β₂ = μ₄ / μ₂². These moment-ratios help determine skewness and kurtosis of a probability distribution.

🔑 Definition — Variance (σ²): σ² = E(X - μ)² = Σ (xᵢ - μ)² f(xᵢ), a measure of dispersion of the probability distribution.

📐 Formula: σ² = E(X²) - [E(X)]² → Variance equals the expected value of X² minus the square of the expected value of X.

Properties of Expected Values

1. If c is a constant, then E(c) = c. The expected value of a constant is the constant itself.

2. If X is a discrete random variable and a and b are constants, then E(aX + b) = a E(X) + b.

🔑 Definition — Linearity of Expectation: For constants a and b, E(aX + b) = a E(X) + b.

📌 Example: For three fair coins, X = number of heads, with distribution: P(0)=1/8, P(1)=3/8, P(2)=3/8, P(3)=1/8. E(X) = 0(1/8) + 1(3/8) + 2(3/8) + 3(1/8) = 12/8 = 1.5. Now find E(2X+3): E(2X+3) = 2E(X) + 3 = 2(1.5) + 3 = 3 + 3 = 6. Verification: For x=0: 2(0)+3=3, P=1/8 → 3/8; x=1: 5, P=3/8 → 15/8; x=2: 7, P=3/8 → 21/8; x=3: 9, P=1/8 → 9/8. Sum = (3+15+21+9)/8 = 48/8 = 6. ✓

⭐ Key Takeaways

The distribution function F(x) for a discrete random variable is a step function that gives cumulative probabilities, and its graph visually shows jumps at each possible value of X. Mathematical expectation E(X) is a weighted average representing the long-run average value of a random variable. Variance σ² = E(X²) - [E(X)]² measures spread, while moments about the origin and about the mean help characterize distributions. The linearity property E(aX + b) = aE(X) + b is crucial for simplifying expectation calculations.

🧠 Quick Revision Questions

1. For the three-coin toss example, what is the probability of getting at most one head?
2. How would you interpret F(2) = 7/8 in the context of the three-coin toss problem?
3. A salesman earns $30 if it rains (P=0.3) and loses $6 if it doesn't (P=0.7). What is his expected daily earnings?
4. Using the shortcut formula, how would you compute the variance of a discrete random variable?
5. If E(X) = 1.5, what is E(2X + 3) and which property of expectation justifies your answer?

📘 Lecture 24 — Chebychev’s Inequality and Concept of Continuous Probability Distribution

📖 Overview: This lecture introduces Chebychev’s Inequality, a fundamental theorem that provides a guaranteed lower bound for the probability that a random variable falls within a specified number of standard deviations from its mean, applicable to any distribution with finite mean and variance. The lecture then transitions to continuous random variables, defining their probability density functions, distribution functions, and computing probabilities through integration, laying the groundwork for statistical inference.

🗂️ Topics Covered

The lecture covers Chebychev’s Inequality with examples using discrete probability distributions, including its application to business ventures. It then introduces the concept of continuous random variables and probability density functions (p.d.f.), their properties, and the distribution function. The lecture concludes with an example of computing probabilities, the distribution function, and conditional probabilities for a continuous distribution.

📝 Lecture Summary

Chebychev’s Inequality

If X is a random variable having mean μ and variance σ² > 0, and k is any positive constant, then the probability that a value of X falls within k standard deviations of the mean is at least 1 – 1/k². Formally: P(μ – kσ < X < μ + kσ) ≥ 1 – 1/k². Alternatively, the probability of observing a value that differs from μ by k or more standard deviations cannot exceed 1/k². This inequality, due to Russian mathematician P.L. Chebychev (1821–1894), holds for all probability distributions having finite mean and variance and provides a means of understanding how the standard deviation measures variability about the mean.

🔑 Definition — Chebychev’s Inequality: For any random variable X with finite mean μ and variance σ²>0, and any positive constant k, P(|X – μ| ≥ kσ) ≤ 1/k², or equivalently P(|X – μ| < kσ) ≥ 1 – 1/k². 📌 Example: For a flower petal distribution with μ = 5.925 and σ = 1.3, using k=2, Chebychev’s inequality states that at least 1 – 1/2² = 1 – 1/4 = 0.75 of the probability lies between μ – 2σ = 3.3 and μ + 2σ = 8.5. The actual probability P(3.3 < X < 8.5) = 0.10 + 0.20 + 0.30 + 0.25 + 0.075 = 0.925, which is greater than 0.75.

📌 Example — Business Ventures: For a distribution of successful ventures out of five with μ = 3.50 and σ = 1.02, the interval μ ± 2σ = (1.46, 5.54). Chebychev’s rule guarantees at least 75% of observations fall in this interval. The actual probability P(2 ≤ X ≤ 5) = 0.132 + 0.309 + 0.360 + 0.168 = 0.969, consistent with both Chebychev’s rule and the Empirical Rule. Fewer than two successful ventures (x=0 or x=1) is unlikely, with exact probability P(x<2) = 0.002 + 0.029 = 0.031.

💡 Why this matters: Chebychev’s Inequality and the Empirical Rule assist in understanding how much data lies between the mean minus a certain number of standard deviations and mean plus that same number of standard deviations, providing a basis for statistical inference.

Concept of Continuous Probability Distribution

A continuous random variable can assume every possible value in an interval [a, b], where a and b may be –∞ and +∞ respectively. The function f(x) is called the probability density function (p.d.f.) of X. A continuous probability distribution has the following properties: i) f(x) ≥ 0, for all x ii) ∫_{-∞}^{∞} f(x) dx = 1 (total area under the curve is 1) iii) The probability that X takes on a value in the interval [c, d] is given by P(c < X < d) = ∫_c^d f(x) dx, which is the area under the curve between X=c and X=d.

Since P(X = k) = ∫_k^k f(x) dx = 0, the probability of a continuous random variable taking any particular value is always zero. Therefore, for a continuous random variable, P(c < X < d), P(c ≤ X < d), P(c < X ≤ d), and P(c ≤ X ≤ d) are regarded as the same.

🔑 Definition — Continuous Random Variable: A random variable X is continuous if its distribution function F(x) is continuous and differentiable everywhere except at isolated points in the given range. The graph of F(x) has no jumps or steps but is a continuous function for all x-values. 📐 Relationship: f(x) = dF(x)/dx, and F(x) = ∫_{-∞}^x f(t) dt, for all x.

Mathematical Expectation, Variance & Moments of a Continuous Probability Distribution

For a continuous random variable X with p.d.f. f(x), the expected value (mean) is defined as E(X) = ∫_{-∞}^{∞} x f(x) dx. Properties of mathematical expectation, such as E(aX + b) = a E(X) + b, hold for continuous distributions. Moments about the origin (μ’_r) are given by μ’r = E(X^r) = ∫{-∞}^{∞} x^r f(x) dx. Moments about the mean (μ_r) are computed using relationships between moments about the origin, from which moment-ratios (β₁ and β₂) are derived.

📌 Example — Expected Value: For f(x) = 2(1 – x), 0 < x < 1, E(X) = 2∫_0^1 x(1 – x) dx = 2[x²/2 – x³/3]₀¹ = 2(1/2 – 1/3) = 1/3. For a=3 and b=5, E(3X+5) = 3(1/3) + 5 = 6, verified by direct integration.

📌 Example — Moments: For f(x) = (3/4)x(2 – x), 0 ≤ x ≤ 2, the first four moments about the origin are: μ’₁ = 1, μ’₂ = 6/5, μ’₃ = 8/5, μ’₄ = 16/7. Moments about the mean: μ₁ = 0, μ₂ = 1/5, μ₃ = 0, μ₄ = 3/35. Moment-ratios: β₁ = μ₃²/μ₂³ = 0 (absolutely symmetric), β₂ = μ₄/μ₂² = 2.14 (platykurtic).

⭐ Key Takeaways

Chebychev’s Inequality provides a universal bound: for any distribution with finite variance, at least 1 – 1/k² of the data lies within k standard deviations of the mean. It works for all distributions, while the Empirical Rule applies specifically to mound-shaped distributions. Continuous random variables have probability density functions where probabilities are areas under the curve, and the probability of any single point is zero. The expected value and moments for continuous distributions are computed using integration, and properties of expectation carry over from discrete cases. The concept of joint (bivariate) probability distributions introduces marginal and conditional probability functions, with independence defined as the joint probability equaling the product of marginal probabilities.

🧠 Quick Revision Questions

1. What does Chebychev’s Inequality state about the probability that a random variable falls within k standard deviations of its mean?
2. For a continuous random variable, why is P(X = k) always zero, and what are the implications for interval probabilities?
3. A p.d.f. is given by f(x) = kx for 0 ≤ x ≤ 2. Find the value of k that makes this a valid density function.
4. For the same p.d.f. f(x) = x/2, 0 ≤ x ≤ 2, compute P(X > 1) and find the distribution function F(x).
5. A joint probability distribution lists f(0,1) = 6/28 and the marginal g(0) = 10/28 and h(1) = 12/28. Are X and Y independent at these values? Why or why not?

📘 Lecture 26 — STA301 Statistics and Probability

📖 Overview: This lecture completes the study of bivariate probability distributions by working through a detailed discrete example involving drawing balls from an urn. It then introduces continuous bivariate distributions, including joint density functions, marginal densities, conditional densities, and the concept of statistical independence. Finally, important properties of expected values for bivariate distributions are presented.

🗂️ Topics Covered

The lecture works through a complete discrete bivariate example finding joint, marginal, and conditional probability distributions and testing independence. It then introduces continuous bivariate distributions with definitions of joint, marginal, and conditional density functions, followed by a worked example. The lecture concludes with two key properties of expected values for sums and products of random variables.

📝 Lecture Summary

BIVARIATE Probability Distributions (Discrete and Continuous)

The lecture begins by completing a discrete example involving drawing 2 balls from an urn containing 3 black, 2 red, and 3 green balls. Let X be the number of black balls and Y be the number of red balls selected.

Using the rule of combinations with the classical definition of probability, the following joint probability distribution table is obtained:

The joint p.d. formula is: f(x,y) = [(3 choose x)(2 choose y)(3 choose 2-x-y)] / (8 choose 2), where x=0,1,2; y=0,1,2; 0 ≤ x+y ≤ 2.

To compute P(X+Y < 1), we identify cells where x+y < 1: (0,0), (0,1), and (1,0). Therefore, P(X+Y < 1) = f(0,0) + f(0,1) + f(1,0) = 3/28 + 6/28 + 9/28 = 18/28 = 9/14.

The marginal p.d. g(x) and h(y) are obtained by summing probabilities across rows and columns respectively:

• g(x): x=0 → 10/28, x=1 → 15/28, x=2 → 3/28
• h(y): y=0 → 15/28, y=1 → 12/28, y=2 → 1/28

The conditional p.d. f(x|1) = P(X=x | Y=1) = f(x,1)/h(1). Since h(1)=12/28=3/7, we compute:

• f(0|1) = (7/3)×(6/28) = 1/2
• f(1|1) = (7/3)×(6/28) = 1/2
• f(2|1) = (7/3)×0 = 0

Hence, the conditional distribution of X given Y=1 is: x=0 → 1/2, x=1 → 1/2, x=2 → 0.

To test independence, check if f(0,1) = g(0)h(1). Since f(0,1)=6/28, g(0)=10/28, and h(1)=12/28, we see that 6/28 ≠ (10/28)×(12/28). Therefore, X and Y are NOT statistically independent.

CONTINUOUS BIVARIATE DISTRIBUTIONS

The bivariate probability density function of continuous r.v.’s X and Y is an integral function f(x,y) satisfying:

🔑 Definition — Bivariate continuous joint p.d.f.: f(x,y) is a function satisfying:

• i) f(x,y) ≥ 0 for all (x,y)
• ii) ∫∫ f(x,y) dx dy = 1 over the entire XY-plane
• iii) P(a ≤ X ≤ b, c ≤ Y ≤ d) = ∫ₐᵇ ∫_cᵈ f(x,y) dy dx

The probability function f(x,y) gives a SURFACE, and computing P(x₁ < X < x₂, y₁ < Y < y₂) means computing the VOLUME under the surface over that rectangular region.

The marginal p.d.f. of continuous r.v. X is: g(x) = ∫{-∞}^{∞} f(x,y) dy, and the marginal p.d.f. of Y is: h(y) = ∫{-∞}^{∞} f(x,y) dx.

The conditional p.d.f. of X given Y=y is: f(x|y) = f(x,y)/h(y), where h(y) > 0. Similarly, f(y|x) = f(x,y)/g(x), where g(x) > 0.

Two continuous r.v.’s X and Y are statistically independent if and only if their joint density can be factorized as: f(x,y) = g(x)h(y) for all possible values of X and Y.

📐 Formula — Continuous bivariate independence condition: f(x,y) = g(x)h(y) for all x, y

An example is worked through with joint p.d.f. f(x,y) = (1/8)(6-x-y), 0 ≤ x ≤ 2, 2 ≤ y ≤ 4, and 0 elsewhere.

a) Verification as a joint density: ∫₀² ∫₂⁴ (1/8)(6-x-y) dy dx = (1/8)[12-4] = 1. Also f(x,y) > 0 in the region. So it satisfies all properties.

b) P(X ≤ 3/2, Y ≤ 5/2) = ∫₀^{3/2} ∫₂^{5/2} (1/8)(6-x-y) dy dx = (1/8)[(15/2)x - x²]_0^{3/2} = (1/8)(45/4 - 9/4) = (1/8)(36/4) = 9/32.

c) Marginal p.d.f. of X: g(x) = ∫₂⁴ (1/8)(6-x-y) dy = (1/4)(3-x), 0 ≤ x ≤ 2. Marginal p.d.f. of Y: h(y) = ∫₀² (1/8)(6-x-y) dx = (1/4)(5-y), 2 ≤ y ≤ 4.

d) Conditional p.d.f. of X given Y=y: f(x|y) = f(x,y)/h(y) = [(1/8)(6-x-y)] / [(1/4)(5-y)] = (6-x-y)/(2(5-y)), 0 ≤ x ≤ 2. Conditional p.d.f. of Y given X=x: f(y|x) = f(x,y)/g(x) = [(1/8)(6-x-y)] / [(1/4)(3-x)] = (6-x-y)/(2(3-x)), 2 ≤ y ≤ 4.

Properties of Expected Values in the case of Bivariate Probability Distributions

🔑 Definition — Property No. 1: E(X+Y) = E(X) + E(Y). This also holds for differences: E(X-Y) = E(X) - E(Y)

🔑 Definition — Property No. 2: If X and Y are independent r.v.’s, then E(XY) = E(X)E(Y)

These properties are valid for both discrete and continuous random variables.

📌 Example: Given the joint distribution:

First find marginal distributions: g(x): x=1→0.25, x=3→0.50, x=5→0.25; h(y): y=2→0.40, y=4→0.60.

E(X) = 1×0.25 + 3×0.50 + 5×0.25 = 0.25 + 1.50 + 1.25 = 3.0 E(Y) = 2×0.40 + 4×0.60 = 0.80 + 2.40 = 3.2 E(X+Y) = E(X) + E(Y) = 3.0 + 3.2 = 6.2

Alternatively, E(X+Y) = ΣΣ(x_i + y_j) f(x_i, y_j) over all i, j. E(XY) = ΣΣ(x_i y_j) f(x_i, y_j) over all i, j.

⭐ Key Takeaways

The joint probability distribution for discrete random variables must sum to 1 over all cells, and the marginal distributions are obtained by summing across rows or columns. Conditional probability distributions are found by dividing joint probabilities by marginal probabilities. For continuous bivariate distributions, volumes under the surface represent probabilities, and marginal densities are found by integrating out the other variable. Two random variables are statistically independent only if their joint distribution equals the product of their marginals for all values. The expected value of a sum always equals the sum of expected values, but the expected value of a product equals the product of expected values only when the variables are independent.

🧠 Quick Revision Questions

1. In the urn example with black, red, and green balls, what is the value of P(X=0, Y=2)?
2. What are the four properties that a continuous bivariate joint density function must satisfy?
3. For the example f(x,y) = (1/8)(6-x-y) with 0≤x≤2, 2≤y≤4, what is P(X>1, Y<3)?
4. When computing the conditional p.d.f. f(x|y), what condition must the marginal p.d.f. h(y) satisfy?
5. If E(X) = 3.0 and E(Y) = 3.2, what is E(2X - 3Y + 5)?

📘 Lecture 27 — Properties of Expected Values in the case of Bivariate Probability Distributions (Detailed discussion) — Covariance & Correlation — Some Well-known Discrete Probability Distributions: — Discrete Uniform Distribution — An Introduction to the Binomial Distribution

📖 Overview: This lecture explores the key properties of mathematical expectation for bivariate probability distributions, establishing foundational rules for expected values of sums and products of random variables. It introduces covariance and correlation as measures of the relationship between two variables, and begins the study of important discrete probability distributions, specifically the discrete uniform and binomial distributions.

🗂️ Topics Covered

The lecture covers the properties of expected values for sums and products of random variables in both discrete and continuous bivariate distributions, verified through detailed examples. It then introduces covariance and the correlation coefficient as measures of linear association, followed by a discussion of the discrete uniform distribution illustrated with a die-tossing example and a lottery scenario. Finally, the lecture provides an introduction to the binomial distribution and the fundamental properties of a binomial experiment.

📝 Lecture Summary

Properties of Expected Values in the case of Bivariate Probability Distributions (Detailed discussion)

This section examines two key properties of mathematical expectation for bivariate distributions. First, the expected value of the sum of two random variables equals the sum of their individual expected values: E(X + Y) = E(X) + E(Y). Second, for independent random variables, the expected value of the product equals the product of their expected values: E(XY) = E(X)E(Y).

🔑 Definition — Mathematical Expectation: For discrete r.v.’s, the expected value of a function h(x,y) is E[h(X,Y)] = ΣᵢΣⱼ h(xᵢ,yⱼ) f(xᵢ,yⱼ).

📐 Formula: E(X + Y) = E(X) + E(Y) → The expected value of the sum is the sum of the expected values.

📐 Formula: E(XY) = E(X)E(Y) → Only holds when X and Y are independent.

📌 Example: For the joint p.d. where X takes values 2 and 4, and Y takes values 1, 3, and 5:

• E(X) = 2 × 0.40 + 4 × 0.60 = 3.2
• E(Y) = 1 × 0.25 + 3 × 0.50 + 5 × 0.25 = 3.0
• E(X + Y) = (2+1)(0.10) + (2+3)(0.20) + (2+5)(0.10) + (4+1)(0.15) + (4+3)(0.30) + (4+5)(0.15) = 6.20 = E(X) + E(Y)
• E(XY) = (2×1)(0.10) + (2×3)(0.20) + (2×5)(0.10) + (4×1)(0.15) + (4×3)(0.30) + (4×5)(0.15) = 9.6 = E(X)E(Y) → confirming independence.

📌 Example: For continuous r.v.’s with joint p.d.f. f(x,y) = x(1+3y²)/4, 0<x<2, 0<y<1:

• E(X) = 4/3, E(Y) = 5/8
• E(X + Y) = 47/24 = E(X) + E(Y) ✓
• E(XY) = 5/6 = E(X)E(Y) ✓

💡 Why this matters: These properties simplify computation and help determine independence between variables.

Covariance & Correlation

This section introduces measures of association between two random variables. Covariance quantifies the direction of the linear relationship, while the correlation coefficient provides a standardized measure of linear correlation strength.

🔑 Definition — Covariance: Cov(X,Y) = E{[X - E(X)][Y - E(Y)]} = E(XY) - E(X)E(Y), measuring how X and Y vary together.

📐 Formula: Cov(X,Y) = E(XY) - E(X)E(Y) → The short-cut formula for covariance.

📐 Formula: ρXY = Cov(X,Y) / (σXσY) → The correlation coefficient, ranging from -1 to 1.

📌 Example: For the joint p.d. with x=0,1,2 and y=0,1,2,3:

• E(X) = 1.10, E(Y) = 1.65
• Var(X) = 0.49, Var(Y) = 0.7275
• E(XY) = 1.90
• Cov(X,Y) = 1.90 - 1.10×1.65 = 0.085
• ρ = 0.085/√(0.49×0.7275) = 0.14 → weak positive linear correlation.

📌 Example: For f(x,y) = x² + xy/3, 0<x<1, 0<y<2:

• E(X) = 13/18, E(Y) = 10/9
• Var(X) = 73/1620, Var(Y) = 26/81
• Cov(X,Y) = -1/162
• ρ = -0.05 → very weak negative linear correlation, almost uncorrelated.

💡 Why this matters: Independent variables have zero covariance/correlation, but the converse is not true—zero correlation does not imply independence.

Some Well-known Discrete Probability Distributions: — Discrete Uniform Distribution

The discrete uniform distribution arises when each possible outcome of a random variable is equally likely to occur. It is characterized by all values having the same probability of occurrence.

🔑 Definition — Discrete Uniform Distribution: A probability distribution where the random variable X assumes each of its k values with equal probability. For a fair die, each value from 1 to 6 has probability 1/6.

📌 Example: Tossing a fair die: X represents the number of dots on the upper-most face. Each X-value from 1 to 6 occurs with probability 1/6. The mean lies at the exact center of the distribution, i.e., E(X) = 3.5.

📌 Example: Lottery with tickets numbered 0000 to 9999, where each number is equally likely to be selected. This reflects the discrete uniform distribution because all outcomes have the same probability of occurrence.

An Introduction to the Binomial Distribution

The binomial distribution is a fundamental discrete probability distribution discovered by James Bernoulli around 1700. It models the number of successes in a fixed number of independent trials with constant success probability.

🔑 Definition — Binomial Distribution: The probability distribution of the number of successes (X) in n independent Bernoulli trials, where each trial has constant success probability p and failure probability q = 1-p.

📌 Example: Tossing a fair coin 5 times, where X represents the number of heads. Key properties include:

• Every toss results in either a head (success) or tail (failure)
• The probability of heads p = 1/2 remains constant
• Each throw is independent
• The total number of tosses n = 5 is fixed in advance

📐 Properties of a Binomial Experiment:

1. Every trial results in a success or a failure
2. Successive trials are independent
3. The probability of success, p, remains constant from trial to trial
4. The number of trials, n, is fixed in advance

⭐ Key Takeaways

The expected value of the sum of two random variables always equals the sum of their individual expected values, regardless of independence. For independent random variables, the expected value of the product equals the product of the expected values. Covariance and correlation provide numerical measures of linear association, with zero values for independent variables, though zero correlation does not guarantee independence. The discrete uniform distribution applies when all outcomes are equally likely, such as with a fair die or a random lottery. The binomial distribution applies to experiments with fixed n independent trials, constant success probability p, and binary outcomes (success/failure).

🧠 Quick Revision Questions

1. What is the relationship between E(X+Y), E(X), and E(Y) in a bivariate probability distribution?
2. Under what condition does E(XY) equal E(X)E(Y)?
3. What is the short-cut formula for covariance, and what does Cov(X,Y) = 0 imply about independence?
4. Define the discrete uniform distribution and give a real-world example where it applies.
5. List the four essential properties of a binomial experiment.

📘 Lecture 28 — Binomial Distribution, Fitting a Binomial Distribution to Real Data, An Introduction to the Hypergeometric Distribution

📖 Overview: This lecture explores the binomial distribution, a fundamental discrete probability distribution for modeling success/failure experiments. It covers its properties, parameters, mean and standard deviation, and shows how to fit a binomial distribution to real-world observed data. The lecture concludes with an introduction to the hypergeometric distribution, which applies when sampling is done without replacement and the probability of success changes on each trial.

🗂️ Topics Covered

The lecture begins by illustrating the binomial distribution using the example of tossing a fair coin five times, detailing its four essential properties and the formula for calculating probabilities. It then presents the formulas for the mean and standard deviation of a binomial distribution and discusses the distribution's shape (symmetrical vs. skewed). The second half of the lecture demonstrates the process of fitting a binomial distribution to real data from a loaded die, including calculating expected frequencies and comparing them to observed frequencies. The lecture concludes by introducing the hypergeometric distribution, its properties, and its probability formula, highlighting when it is used instead of the binomial distribution.

📝 Lecture Summary

Binomial Distribution

The binomial distribution is a very important discrete probability distribution. It applies to a binomial experiment which has four basic and vitally important properties: (1) every trial results in either a success or failure, (2) the probability of success (p) remains constant for all trials, (3) the trials are independent, and (4) the number of trials (n) is fixed in advance.

The probability distribution is given by the formula: P(X = x) = (n choose x) p^x q^(n-x)

Where: n = the total number of trials p = probability of success in each trial q = probability of failure in each trial (i.e., q = 1 - p) x = number of successes in n trials (x = 0, 1, 2, ..., n)

The binomial distribution has two parameters, n and p.

🔑 Definition — Binomial Distribution: A discrete probability distribution that models the number of successes in a fixed number of independent trials, each with the same probability of success.

📐 Formula: P(X = x) = (n chooses x) p^x q^(n-x) → The probability of getting exactly x successes in n trials equals the number of ways to choose x successes from n trials, multiplied by the probability of x successes and the probability of (n-x) failures.

📌 Example: Tossing a fair coin 5 times (n=5, p=0.5). Calculate P(X = 0 heads) = (5 choose 0) (1/2)^0 (1/2)^5 = 1 * 1 * 1/32 = 1/32. P(X = 1 head) = (5 choose 1) (1/2)^1 (1/2)^4 = 5 * (1/2) * (1/16) = 5/32. The complete distribution is: P(0)=1/32, P(1)=5/32, P(2)=10/32, P(3)=10/32, P(4)=5/32, P(5)=1/32.

💡 Why this matters: The formula allows us to calculate the probability of any number of successes in a binomial experiment without having to list all possible outcomes.

Mean and Standard Deviation of the Binomial Distribution

For a binomial distribution, there are simple formulas for the mean and standard deviation. It has been mathematically proved that:

For a binomial distribution: E(X) = np Var(X) = npq so that S.D.(X) = √(npq)

🔑 Definition — Mean of Binomial Distribution (μ): np → The average number of successes expected over many repetitions of the binomial experiment.

🔑 Definition — Standard Deviation of Binomial Distribution (σ): √(npq) → A measure of the spread or variability of the possible number of successes around the mean.

📐 Formulas: μ = np; σ = √(npq) → The mean is the product of trials and success probability; the standard deviation is the square root of the product of trials, success probability, and failure probability.

📌 Example: For the fair coin toss example (n=5, p=0.5, q=0.5): Mean = np = 5(0.5) = 2.5 S.D.(X) = √(5 * 0.5 * 0.5) = √(5/4) = 1.12 This means that if 5 fair coins are tossed an infinite number of times, we should expect to get 2.5 heads on average. The standard deviation of 1.12 gives a measure of the possible variability in the number of heads. Coefficient of Variation: C.V. = σ/μ * 100 = 1.12/2.5 * 100 = 44.8%

Shape of the Binomial Distribution

The binomial distribution is not always symmetrical. Its shape depends on the values of p and q:

• It is symmetrical only when p = q = ½.
• It is skewed to the right if p < q.
• It is skewed to the left if p > q.

The degree of skewness (or asymmetry) decreases as n increases.

🔑 Definition — Skewness: The asymmetry in a probability distribution where the tail on one side is longer or fatter than the other. For binomial, skewness depends on the relative values of p and q.

Fitting a Binomial Distribution to Real Data

When we have observed data from a binomial experiment, we can fit a theoretical binomial distribution to it. This involves estimating the parameters n and p from the data.

The procedure is:

1. Determine n from the largest x-value in the data.
2. Estimate p using the relationship μ = np, but since μ is unknown, we replace it with its estimate x̄ (sample mean). So, x̄ = np.
3. Calculate the theoretical probabilities using the estimated binomial distribution.
4. Multiply these probabilities by the total frequency to get expected frequencies.
5. Compare observed frequencies with expected frequencies to assess the goodness of fit.

📌 Example: The data below was obtained by tossing a loaded die 5 times, noting the number of sixes:

Solution:

1. n = 5 (largest x-value)
2. Calculate x̄: x̄ = Σ(fi xi) / Σfi = (0 + 56 + 148 + 117 + 72 + 5) / 200 = 398/200 = 1.99
3. Using x̄ = np: 5p = 1.99 → p = 0.398
4. The fitted binomial distribution is b(x; 5, 0.398)
5. Calculate probabilities and expected frequencies:

Comparing observed and expected frequencies shows not much discrepancy, indicating a reasonably good fit. The Chi-Square Test of Goodness of Fit is a formal method to determine if the fit is adequate.

💡 Why this matters: Fitting a theoretical distribution to real data allows us to model and analyze real-world phenomena using a known mathematical framework.

Real-Life Application from Industry

The binomial distribution has many real-life applications, such as quality control. When dealing with a success/failure situation, we are often dealing with a binomial experiment. Examples include proportion of smokers, literates, or males in a population.

📌 Example: A factory has a past record of 7% defective articles. A new law states there should not be more than 5% defective. The factory owner claims defectives have decreased. An inspector examines a sample of 20 items. What is the probability of finding 2 or more defective items (so a fine is imposed)?

Solution: This is a binomial experiment (n=20, p=0.07, q=0.93). P(X ≥ 2) = 1 - P(X < 2) = 1 - [P(X=0) + P(X=1)] P(X=0) = (20 choose 0)(0.07)^0(0.93)^20 = 1 * 1 * 0.234 = 0.234 P(X=1) = (20 choose 1)(0.07)^1(0.93)^19 = 20 * 0.07 * 0.353 = 0.4942 (correcting the lecture's calculation to 0.353 * 1.4? wait, let's do carefully) P(X=1) = 20 * 0.07 * (0.93)^19 ≈ 1.4 * 0.353 ≈ 0.494? Wait, 20*0.07 = 1.4. (0.93)^19 can be calculated, but lecture says 0.353 for P(X=1) and 0.234 for P(X=0). Let's use lecture's numbers. P(X=0) = 0.234 P(X=1) = 20 * 0.07 * 0.93^19 = 1.4 * 0.93^19. Lecture says this equals 0.353. So, P(X=1) = 0.353. P(X ≥ 2) = 1 - (0.234 + 0.353) = 1 - 0.587 = 0.413 = 41.3%

Hence, there is a substantial probability (more than 40%) that the inspector will find two or more defective articles, meaning a considerable chance the factory will be fined.

Hypergeometric Probability Distribution

There are many experiments where the condition of independence is violated and the probability of success does not remain constant for all trials. Such experiments are called hypergeometric experiments.

Properties of a Hypergeometric Experiment:

1. The outcomes of each trial may be classified into one of two categories, success and failure.
2. The probability of success changes on each trial.
3. The successive trials are not independent.
4. The experiment is repeated a fixed number of times.

The number of successes, X, in a hypergeometric experiment is called a hypergeometric random variable, and its probability distribution is the hypergeometric distribution.

The formula is: P(X = x) = ( (k choose x) * (N-k choose n-x) ) / (N choose n)

Where: N = number of units in the population n = number of units in the sample k = number of successes in the population

The hypergeometric probability distribution has three parameters: N, n, and k.

It is appropriate when:

• A random sample of size n is drawn WITHOUT REPLACEMENT from a finite population of N units.
• k of the units are of one kind (classified as success) and the remaining N-k of another kind (classified as failure).

🔑 Definition — Hypergeometric Distribution: A discrete probability distribution that models the number of successes in a sample drawn without replacement from a finite population containing a fixed number of successes.

📐 Formula: P(X = x) = [C(k, x) * C(N-k, n-x)] / C(N, n) → The probability of getting exactly x successes in a sample of size n is the number of ways to choose x successes from k successes multiplied by the number of ways to choose (n-x) failures from (N-k) failures, divided by the total number of ways to choose a sample of size n from the population of size N.

💡 Why this matters: The hypergeometric distribution is used when sampling is done without replacement, where the binomial distribution would not be appropriate because the probability of success changes with each draw.

⭐ Key Takeaways

1. The binomial distribution is defined by two parameters, n (number of trials) and p (probability of success), and its formula P(X=x) = (n choose x) p^x q^(n-x) allows for calculating probabilities of any number of successes.
2. For any binomial distribution, the mean is np and the standard deviation is √(npq), and its shape is symmetrical only when p=q=0.5, skewed right when p<q, and skewed left when p>q.
3. To fit a binomial distribution to real data, estimate p using the sample mean (x̄ = np) and then calculate expected frequencies to compare with observed frequencies.
4. The hypergeometric distribution is used instead of the binomial distribution when sampling is done without replacement from a finite population, making trials dependent and changing the probability of success.
5. The hypergeometric probability formula uses three parameters (N, n, k) and is expressed as P(X=x) = [(C(k,x) * C(N-k, n-x)] / C(N,n).

🧠 Quick Revision Questions

1. What are the four properties a random experiment must satisfy to be considered a binomial experiment?
2. A binomial distribution has n=10 and p=0.3. Calculate its mean and standard deviation.
3. Explain the procedure for fitting a binomial distribution to a given set of observed frequencies.
4. In the loaded die example with n=5 and total frequency 200, why was the calculated value of p=0.398 considered evidence that the die was not fair?
5. Under what specific conditions would you use the hypergeometric distribution instead of the binomial distribution?

📘 Lecture 29 — Hypergeometric Distribution, Poisson Distribution, Limiting Approximation to the Binomial, Poisson Process, Continuous Uniform Distribution

📖 Overview: This lecture completes the discussion of the hypergeometric probability distribution and introduces the Poisson distribution as both a limiting approximation to the binomial and a distribution in its own right. It also begins the discussion of continuous probability distributions with the uniform distribution. Understanding these distributions is essential for modeling real-world phenomena where events occur randomly over time or space.

🗂️ Topics Covered

The lecture covers hypergeometric distribution properties and its relationship to the binomial distribution, the Poisson distribution as a limiting approximation to the binomial when n is large and p is small, the Poisson process for random events over intervals, properties of the Poisson distribution including its mean-variance equality, fitting a Poisson distribution to real data, and the continuous uniform distribution with its mean, variance, and rectangular shape.

📝 Lecture Summary

Properties of the Hypergeometric Experiment

A hypergeometric experiment has the following properties: The outcomes of each trial may be classified into one of two categories, success and failure. The probability of success changes on each trial. The successive trials are not independent. The experiment is repeated a fixed number of times.

The number of successes, X, in a hypergeometric experiment is called a hypergeometric random variable, and its probability distribution is called the hypergeometric distribution.

🔑 Definition — Hypergeometric Probability Distribution: When the hypergeometric random variable X assumes a value x, the hypergeometric probability distribution is given by:

📐 Formula: P(X = x) = [C(k, x) × C(N-k, n-x)] / C(N, n)

Where:

• N = number of units in the population
• n = number of units in the sample
• k = number of successes in the population

The hypergeometric probability distribution has three parameters: N, n, and k.

The hypergeometric probability distribution is appropriate when:

• A random sample of size n is drawn WITHOUT REPLACEMENT from a finite population of N units
• k of the units are of one kind (classified as success) and the remaining N-k of another kind (classified as failure)

📌 Example: The names of 5 men and 5 women are written on slips of paper and placed in a hat. Four names are drawn. What is the probability that 2 are men and 2 are women?

Let us regard 'men' as success. Then X will denote the number of men. We have:

• N = 5 + 5 = 10 names to be drawn from
• n = 4 (drawing a sample of size 4 out of a 'population' of size 10)
• k = 5 (since there are 5 men in the population of 10)
• Possible values of X are 0, 1, 2, 3, 4

The hypergeometric distribution is given by: P(X = x) = [C(5, x) × C(5, 4-x)] / C(10, 4)

The required probability P(X = 2) is: P(X = 2) = [C(5, 2) × C(5, 2)] / C(10, 4) = (10 × 10) / 210 = 100/210 = 10/21

In other words, the probability is a little less than 50% that two of the four names drawn will be those of men.

Properties of the Hypergeometric Distribution

The mean and variance of the hypergeometric probability distribution are:

📐 Formulas:

• Mean: μ = n × (k/N)
• Variance: σ² = n × (k/N) × ((N-k)/N) × ((N-n)/(N-1))

If N becomes indefinitely large, the hypergeometric probability distribution tends to the BINOMIAL probability distribution.

Sampling and Distribution Selection:

• When sampling with replacement, the population remains undisturbed. The various trials (draws) are independent, and the binomial formula can be used.
• When sampling without replacement from a finite population, the constitution of the population changes at every draw. The trials are not independent, and the hypergeometric formula is appropriate.
• If the population size is much larger than the sample size (regarded as 'infinite' population), even without replacement, the population remains almost undisturbed. The trials can be assumed independent, and the binomial formula can be applied.

Rule of Thumb: The binomial formula can be applied when drawing a sample from a finite population without replacement and the sample size n is not more than 5 percent of the population size N (i.e., n < 0.05N). When n is greater than 5 percent of N, the hypergeometric formula should be used.

💡 Why this matters: This rule helps determine which probability distribution to use in real sampling situations, balancing accuracy with computational simplicity.

Poisson Distribution

The Poisson distribution is named after the French mathematician Siméon Denis Poisson (1781-1840). It arises in two situations:

1. As a limiting approximation to the binomial distribution when p, the probability of success, is very small but n, the number of trials, is so large that the product np = μ is of a moderate size.
2. As a distribution in its own right by considering a Poisson Process where events occur randomly over a specified interval of time or space or length.

Such random events might be:

• Number of typing errors per page in a book
• Number of traffic accidents in a particular city in a 24-hour period
• Number of taxicab arrivals at an intersection per day
• Number of radioactive particles emitted in a given period
• Number of flaws per unit length of some material

🔑 Definition — Poisson Distribution (Limiting Form): If n goes to infinity and p approaches zero in such a way that μ = np remains constant, then the limiting form of the binomial probability distribution is:

📐 Formula: P(X = x) = (e^(-μ) × μ^x) / x!, where x = 0, 1, 2, ..., ∞ and e = 2.71828

The Poisson distribution has only one parameter μ > 0. The parameter μ may be interpreted as the mean of the distribution.

Practical Rule: Most statisticians use the Poisson approximation to the binomial when p is 0.05 or less and n is 20 or more. The larger n is and the smaller p is, the better the approximation.

📌 Example: Two hundred passengers have made reservations for an airplane flight. If the probability that a passenger who has a reservation will not show up is 0.01, what is the probability that exactly three will not show up?

Solution: This is a binomial experiment with n = 200 and p = 0.01. Since p is very small and n is considerably large, apply the Poisson distribution using μ = np = (200)(0.01) = 2.

If X represents the number of successes (not showing up): P(X = 3) = (e^(-2) × 2³) / 3!

e^(-2) = 1/(2.71828)² = 0.1353

P(X = 3) = (0.1353 × 8) / (3 × 2 × 1) = 1.0824 / 6 = 0.1804

Poisson Process

A Poisson Process may be defined as a physical process governed at least in part by some random mechanism. Stated differently, a Poisson process represents a situation where events occur randomly over a specified interval of time or space or length.

🔑 Definition — Poisson Process Formula: The formula valid in the case of a Poisson process is:

📐 Formula: P(X = x) = (e^(-λt) × (λt)^x) / x!

Where:

• λ = average number of occurrences of the outcome of interest per unit of time
• t = number of time-units under consideration
• x = number of occurrences of the outcome of interest in t units of time

📌 Example: Telephone calls are being placed through a certain exchange at random times on the average of four per minute. Assuming a Poisson Process, determine the probability that in a 15-second interval, there are 3 or more calls.

Solution:

Step 1: Identify the unit of time — take a minute as the unit of time.

Step 2: Identify λ — the average number of occurrences per unit of time. λ = 4 calls per minute

Step 3: Identify t — the number of time-units under consideration. 15 seconds = 15/60 = 1/4 minutes, therefore t = 1/4

Step 4: Compute λt: λt = 4 × 1/4 = 1

Step 5: Apply the Poisson formula: P(X = x) = (e^(-λt) × (λt)^x) / x!

Since λt = 1, we need P(X ≥ 3) in a 15-second interval.

P(X ≥ 3) = 1 - P(X < 3) = 1 - [P(X=0) + P(X=1) + P(X=2)] = 1 - Σ(x=0 to 2) [e^(-1) × (1)^x] / x! = 1 - Σ(x=0 to 2) (0.3679 × 1^x) / x! [e^(-1) = 0.3679]

= 1 - [0.3679/0! + 0.3679/1! + 0.3679/2!] = 1 - [0.3679 + 0.3679 + 0.18395] = 1 - 0.91975 = 0.08025

Hence the probability is only 8% (very low probability) that in a 15-second interval, the telephone exchange receives 3 or more calls.

Properties of the Poisson Distribution

🔑 Main properties:

• Mean equals Variance: If the random variable X has a Poisson distribution with parameter μ, then:

  • E(X) = μ
  • Var(X) = μ

• Shape: The shape of the Poisson distribution is positively skewed. The distribution tends to be symmetrical as μ becomes larger and larger.

Comparison with Binomial: Whereas the binomial distribution can be symmetric, positively skewed, or negatively skewed (depending on whether p = 1/2, p < 1/2, or p > 1/2), the Poisson distribution can never be negatively skewed.

Fitting a Poisson Distribution to Real Data

The procedure is very similar to fitting the binomial distribution:

1. The population mean μ is replaced by the sample mean X̄
2. The probabilities of the various values of X are computed using the Poisson formula
3. The chi-square test of goodness of fit enables us to determine whether or not it is a good fit (whether the discrepancy between expected frequencies and observed frequencies is small)

Important Mathematical Points:

1. The Poisson approximation to the binomial formula works well when n > 20 and p < 0.05
2. When the Poisson is used to approximate the binomial which is being used to approximate the hypergeometric, the rule of thumb is that the Poisson distribution can be used to approximate the hypergeometric distribution when n < 0.05N, n > 20, and p < 0.05

Uniform Distribution

A random variable X is said to be uniformly distributed if its density function is defined as:

🔑 Definition — Uniform Distribution (Rectangular Distribution):

📐 Formula: f(x) = 1/(b - a), for a ≤ x ≤ b

The graph of this distribution is a rectangle with:

• Base = (b - a)
• Height = 1/(b - a)

Why it is a proper probability density function:

1. Since a < b, therefore f(x) > 0
2. ∫(-∞ to ∞) f(x) dx = ∫(a to b) 1/(b-a) dx = [x/(b-a)](a to b) = (b-a)/(b-a) = 1

The total area can also be obtained from the simple formula: Area of rectangle = (Base) × (Height) = (b - a) × 1/(b - a) = 1

🔑 Properties — Uniform Distribution:

• Mean: μ = (a + b)/2
• Variance: σ² = (b - a)²/12

The distribution derives its name from the fact that its density is constant or uniform over the interval [a, b] and is 0 elsewhere. It is also called the rectangular distribution because its total probability is confined to a rectangular region.

Characteristics: The uniform probability distribution provides a model for continuous random variables that are evenly distributed over a certain interval. A uniform random variable is just as likely to assume a value in one interval as in any other interval of equal size. There is no clustering of values around any value — instead, there is an even spread over the entire region of possible values.

📌 Example 1: If a short exists in a 5-meter stretch of electrical wire, it may have an equal probability of being in any particular 1-centimeter segment along the line.

📌 Example 2: If a safety inspector plans to choose a time at random during the 4 afternoon work-hours to pay a surprise visit, then each 1-minute time-interval in this 4 work-hour period will have an equally likely chance of being selected for the visit.

The uniform distribution also arises in the study of rounding off errors.

⭐ Key Takeaways

The hypergeometric distribution is the correct model when sampling without replacement from a finite population, with the binomial being applicable only when the sample size is less than 5% of the population. The Poisson distribution serves as both a limiting approximation to the binomial when n is large (≥20) and p is small (≤0.05), and as a distribution for modeling random events occurring over time or space through the Poisson process. A defining characteristic of the Poisson distribution is that its mean equals its variance, and it is always positively skewed. The continuous uniform distribution represents variables with constant probability over an interval, with mean (a+b)/2 and variance (b-a)²/12.

🧠 Quick Revision Questions

1. What are the four properties of a hypergeometric experiment, and how do they differ from those of a binomial experiment?

2. Under what condition can the binomial distribution be used as an approximation to the hypergeometric distribution when sampling without replacement?

3. Calculate the probability using the Poisson distribution: If a bookstore averages 2 customer inquiries per hour, what is the probability of exactly 4 inquiries in a 2-hour period?

4. What is the relationship between the mean and variance of a Poisson distribution, and how does its skewness change as μ increases?

5. A bus arrives at a stop every 10 minutes (uniformly distributed). What is the probability that a randomly arriving passenger waits more than 7 minutes?

📘 Lecture 30 — Normal Distribution

📖 Overview: This lecture provides a comprehensive introduction to the normal distribution, its mathematical definition, key properties, and the standard normal distribution. It covers how to use the area table for direct and inverse probability calculations and concludes with the normal approximation to the binomial distribution for large samples. Understanding this lecture is crucial for performing probability calculations for continuous data and for approximating discrete distributions.

🗂️ Topics Covered

The lecture begins with the mathematical definition of the normal distribution and its important properties, including symmetry, asymptotic nature, and fixed area proportions. It then introduces the standard normal distribution and the process of standardization. The direct use of the area table is demonstrated through an example of dishwasher lifespan, followed by the inverse use of the table to find a minimum acceptable height for police applicants. The lecture concludes with the normal approximation to the binomial distribution, including the continuity correction, illustrated by a malaria death rate example.

📝 Lecture Summary

Normal Distribution

A continuous random variable X is normally distributed with mean μ and standard deviation σ if its probability density function is given by:

f(x) = 1/(σ√(2π)) * e^{-1/2 * [(x-μ)/σ]^2}, for -∞ < x < ∞

where π ≈ 3.1416 and e ≈ 2.71828. This formula defines a family of distributions depending on the two parameters μ and σ.

🔑 Definition — Normal Distribution: A continuous probability distribution that is symmetric about its mean, bell-shaped, and defined by its mean (μ) and standard deviation (σ).

📐 Formula: f(x) = 1/(σ√(2π)) * e^{-1/2 * [(x-μ)/σ]^2} → This is the probability density function for a normal random variable X with mean μ and standard deviation σ.

Important Properties of the Normal Distribution

Property 1: For N(μ, σ^2), μ is the mean and σ is the standard deviation. A change in μ shifts the distribution left or right along the x-axis, while a change in σ determines the flatness or peakedness of the curve.

Property 2: The normal curve is asymptotic to the x-axis as x → ±∞.

Property 3: Because of symmetry, 50% of the area is to the right of the mean (μ), and 50% is to the left.

Property 4: The density function attains its maximum value at x = μ, so the mean, median, and mode are all equal to μ.

Property 5: Since the normal distribution is absolutely symmetrical, the third moment about the mean, μ₃ = 0.

Property 6: For the normal distribution, it can be mathematically proved that μ₄ = 3σ⁴.

Property 7: The moment ratios of the normal distribution are β₁ = 0 and β₂ = 3. Since β₂ = 3, this value is a criterion for measuring kurtosis. If β₂ > 3, the distribution is leptokurtic (more peaked); if β₂ < 3, it is platykurtic (less peaked).

Property 8: Areas under the normal curve remain in fixed proportions within a specified number of standard deviations on either side of μ:

• The interval μ ± 1σ always contains 68.26% of the total area.
• The interval μ ± 2σ always contains 95.44% of the total area.
• The interval μ ± 3σ always contains 99.73% of the total area.

💡 Why this matters: This property is the theoretical foundation for the Empirical Rule discussed in descriptive statistics, which states that for approximately symmetric hump-shaped distributions, about 68% of data lies within ±1 standard deviation, 95% within ±2, and nearly 100% within ±3.

Property 9: The normal curve contains points of inflection where the direction of concavity changes. Their coordinates are (μ - σ, 1/(σ√(2πe))) and (μ + σ, 1/(σ√(2πe))).

The Standard Normal Distribution

A normal distribution with mean 0 and standard deviation 1 is known as the standard normal distribution. Areas under the normal curve have been computed for this specific distribution using numerical integration, and these values are available in the Area Table of the Standard Normal Curve.

🔑 Definition — Standard Normal Distribution: A normal distribution whose mean is zero (μ = 0) and whose standard deviation is one (σ = 1), denoted as N(0,1).

The Process of Standardization

The formula for converting a normal distribution N(μ, σ) to the standard normal distribution N(0, 1) is:

Z = (X - μ) / σ

📐 Formula: Z = (X - μ) / σ → This formula converts any value X from a normal distribution with mean μ and standard deviation σ into a Z-score, which represents the number of standard deviations X is from the mean in the standard normal distribution.

Direct Use of the Area Table

📌 Example: The length of life for an automatic dishwasher is approximately normally distributed with a mean life of 3.5 years and a standard deviation of 1.0 years. If this type of dishwasher is guaranteed for 12 months, what fraction of sales will require replacement?

Step 1: Standardize the warranty period (X = 1.0 year): Z = (1.0 - 3.5) / 1.0 = -2.5 / 1 = -2.5

Step 2: We need the area under the standard normal curve from -∞ to Z = -2.5. From the area table, the area from 0 to 2.5 is 0.4938.

Step 3: The area from Z = 2.5 to ∞ is 0.5 - 0.4938 = 0.0062.

Step 4: Due to symmetry, the area from -∞ to -2.5 is also 0.0062.

Therefore, the probability of a dishwasher lasting less than a year is 0.0062 (0.62%).

Inverse Use of the Area Table

📌 Example: The heights of applicants to the police force are normally distributed with mean 170 cm and standard deviation 3.8 cm. If the shortest 30% of applicants are to be rejected, what is the minimum acceptable height?

Step 1: We need the Z-value to the left of which 30% of the area lies. From the area table, the Z-value corresponding to an area of 0.2000 (the area from 0 to that Z) is 0.52 (since area from 0 to 0.52 is 0.1985, which is closest to 0.2000).

Step 2: Since we want the lower 30%, the Z-value is negative: Z = -0.52.

Step 3: The standardization formula can be rewritten as: X = μ + σZ. Substituting: X = 170 + 3.8(-0.52) = 170 - 1.976 = 168.024 cm.

Therefore, the minimum acceptable height is approximately 168 cm.

Normal Approximation to the Binomial Distribution

When n is large and neither p nor q is close to zero, the binomial distribution can be closely approximated by the normal distribution.

🔑 Definition — Normal Approximation to the Binomial: A method used to compute binomial probabilities by using the normal distribution when n is sufficiently large, typically when both np ≥ 5 and nq ≥ 5.

Key Points for Application:

1. The mean and variance of the normal distribution used for approximation are the mean and variance of the binomial distribution: μ = np and σ² = npq.
2. A continuity correction must be applied because the binomial distribution is discrete and the normal is continuous. Each integer value x is replaced by the interval x - 0.5 to x + 0.5.

📌 Example: The death rate from Malaria in a province is 20%. Find the probability that the number of deaths is between 70 and 80 (inclusive) out of 500 patients.

Step 1: Check conditions: np = 500 * 0.20 = 100 ≥ 5 and nq = 500 * 0.80 = 400 ≥ 5, so we can use the normal approximation.

Step 2: Calculate mean and standard deviation: μ = np = 100, σ² = npq = 500 * 0.20 * 0.80 = 80, so σ = √80 ≈ 8.94.

Step 3: Apply continuity correction. P(70 ≤ X ≤ 80) is replaced by P(69.5 < X < 80.5).

Step 4: Standardize the values and compute the area under the standard normal curve between them. The computed probability is approximately 0.0145 (1.5%).

💡 Why this matters: This technique allows us to calculate probabilities for large binomial samples without cumbersome calculations, as long as the binomial distribution is approximately symmetric (when p is close to 0.5 and n is large).

⭐ Key Takeaways

The normal distribution is a symmetric, bell-shaped continuous distribution completely defined by its mean (μ) and standard deviation (σ). For any normal distribution, fixed proportions of the total area lie within 1, 2, and 3 standard deviations of the mean (68.26%, 95.44%, and 99.73%, respectively). All problems involving normal distributions can be solved by standardizing to the standard normal distribution (Z = (X - μ)/σ) and using the area table. The normal distribution can also approximate the binomial distribution when both np and nq are at least 5, but a continuity correction (adding/subtracting 0.5) must be applied because the binomial is discrete while the normal is continuous.

🧠 Quick Revision Questions

1. What is the formula for the probability density function of the normal distribution, and what do μ and σ represent?
2. List three key properties of the normal distribution related to symmetry, asymptotes, and fixed area proportions.
3. What is the standard normal distribution, and what is the formula for standardizing any normal distribution?
4. When can the normal distribution be used to approximate the binomial distribution, and what is the continuity correction and why is it necessary?
5. In the inverse use of the area table, how do you find an X-value when given a specific area (probability) under the normal curve?

📘 Lecture 31 — Sampling Distribution of X̄; Central Limit Theorem

📖 Overview: This lecture introduces the concept of sampling distributions, a cornerstone of statistical inference. It explains how the sample mean, X̄, behaves across repeated samples from a population, leading to the Central Limit Theorem—a fundamental result that underpins both estimation and hypothesis testing.

🗂️ Topics Covered

The lecture covers inferential statistics branches, the definition and construction of sampling distributions of X̄, mean and standard deviation (standard error) of this distribution, and the Central Limit Theorem. It provides a detailed illustration using all possible samples of sizes 2, 3, and 4 from a rectangular population to reveal properties of the sampling distribution and its convergence to normality.

📝 Lecture Summary

INFERENTIAL STATISTICS

That branch of Statistics which enables us to draw conclusions or inferences about various phenomena on the basis of real data collected on sample basis. In this regard, the first point to be noted is that statistical inference can be divided into two main branches--- estimation, and hypothesis-testing. Estimation itself can be further divided into two branches --- point estimation and interval estimation. The second important point is that the concept of sampling distributions forms the basis for both estimation and hypothesis-testing.

SAMPLING DISTRIBUTION

The probability distribution of any statistic (such as the mean, the standard deviation, the proportion of successes in a sample, etc.) is known as its sampling distribution. In this regard, the first point to be noted is that there are two ways of sampling --- sampling with replacement, and sampling without replacement. In case of a finite population containing N elements, the total number of possible samples of size n that can be drawn from this population with replacement is N^n. In case of a finite population containing N elements, the total number of possible samples of size n that can be drawn from this population without replacement is NCn.

We illustrate the concept of the sampling distribution of X̄ with the help of the following example:

🔑 Definition — Sampling Distribution of X̄: The probability distribution of the sample mean, X̄, obtained by considering all possible samples of a given size from a population.

EXAMPLE: Let X denotes the number of faults in a car. Then X can take the values 0, 1, 2, 3, and 4, the probability of each of these X values is 1/5. Hence, we have the following rectangular (uniform) probability distribution:

MEAN AND VARIANCE OF THE POPULATION DISTRIBUTION

μ = E(X) = Σ x f(x) = 2 σ² = Var(X) = E(X²) – [E(X)]² = Σ x² f(x) – [Σ x f(x)]² = 6 – 2² = 6 – 4 = 2

Practically speaking, only a sample of the cars will be tested at any one occasion, and as such, we are interested in considering the results that would be obtained if a sample of vehicles is tested. Let us consider the situation when only two cars are tested after being selected at the roadside by a mobile testing station.

The situation is equivalent to drawing all possible samples of size 2 from this probability distribution (i.e. the population) WITH REPLACEMENT. There are 5² = 25 possible samples. The table of sample means is produced. Some possible sample means occur several times. A frequency distribution of the sample means is constructed.

📌 Example: Sampling Distribution of X̄ for n=2

This is the SAMPLING DISTRIBUTION of the mean. Its graph is triangular and symmetric.

MEAN AND VARIANCE OF THE SAMPLING DISTRIBUTION OF X̄

The mean is given by μx̄ = E(X̄) = Σ x̄ f(x̄) and the variance by σ²x̄ = Var(X̄) = E(X̄²) – [E(X̄)]² = Σ x̄² f(x̄) – [Σ x̄ f(x̄)]². The square root of the variance is the standard deviation, and the standard deviation of a sampling distribution is termed as its standard error.

Computing for the n=2 case: μx̄ = Σ x̄ f(x̄) = 50/25 = 2 σ²x̄ = E(X̄²) – [E(X̄)]² = 5 – 2² = 5 – 4 = 1 σx̄ = √1 = 1

These computations lead to two very important properties:

Property No. 1 In the case of sampling with replacement as well as in the case of sampling without replacement, we have: μx̄ = μ In this example: μ = 2, Hence μx̄ = 2. The mean of the sampling distribution of X̄ is equal to the population mean.

Property No. 2 In case of sampling with replacement: σx̄ = σ / √n In this example: σ = √2, σx̄ = √2 / √2 = 1. The standard error of the sampling distribution of X̄ is equal to the population standard deviation divided by the square root of the sample size.

In case of sampling without replacement from a finite population: σx̄ = (σ / √n) * √((N – n) / (N – 1))

The factor √((N – n) / (N – 1)) is known as the finite population correction (fpc). If the sample size n is much smaller than the population size N, the fpc is approximately equal to 1, and as such, is not required. In sampling from a finite population, we apply the fpc only if the sample size is greater than 5% of the population size.

Next, we consider the shape of the sampling distribution of X̄. As indicated by the line chart, the above sampling distribution for n=2 is absolutely symmetric and triangular. What happens to the shape with increased sample size?

SAMPLING DISTRIBUTION FOR SAMPLES OF SIZE 3 If all possible samples of size 3 were taken (5³ = 125 samples), the sampling distribution has a more bell-shaped appearance.

SAMPLING DISTRIBUTION FOR SAMPLES OF SIZE 4 If all possible samples of size 4 were taken (5⁴ = 625 samples), the sampling distribution becomes even closer to a bell-shaped curve.

In all three cases the line charts are symmetrical, but as the sample size increases, the overall configuration changes from a triangular distribution to a bell-shaped distribution. When relatively large samples are taken, this bell-shaped distribution assumes the form of a ‘normal’ distribution (also called the ‘Gaussian’ distribution), and this happens irrespective of the form of the parent population. (In this problem, the population of defective items in a car is rectangular.)

CENTRAL LIMIT THEOREM

The theorem states that: “If a variable X from a population has mean μ and finite variance σ², then the sampling distribution of the sample mean X̄ approaches a normal distribution with mean μ and variance σ²/n as the sample size n approaches infinity.” As n → ∞, the sampling distribution of X̄ approaches normality.

Due to the Central Limit Theorem, the normal distribution has found a central place in the theory of statistical inference. Since, in many situations, the sample is large enough for our sampling distribution to be approximately normal, therefore we can utilize the mathematical properties of the normal distribution to draw inferences about the variable of interest. The rule of thumb in this regard is that if the sample size, n, is greater than or equal to 30, then we can assume that the sampling distribution of X̄ is approximately normally distributed. On the other hand, if the POPULATION sampled is normally distributed, then the sampling distribution of X̄ will also be normal regardless of sample size. In other words, X̄ will be normally distributed with mean μ and variance σ²/n.

💡 Why this matters: The Central Limit Theorem is what allows statisticians to use the normal distribution to make inferences about population parameters from sample statistics, even when the original population is not normally shaped. It is the theoretical foundation for confidence intervals and hypothesis tests about the mean.

⭐ Key Takeaways

The sampling distribution of the sample mean is the probability distribution of all possible sample means from a population. Its mean always equals the population mean (μx̄ = μ). Its standard deviation, called the standard error, equals σ/√n for sampling with replacement, and with a finite population correction factor when sampling without replacement. The Central Limit Theorem states that for large samples (n ≥ 30), the sampling distribution of X̄ is approximately normal regardless of the population's shape, and if the population itself is normal, X̄ is normal for any sample size. This theorem is the bedrock of inferential statistics, enabling valid probability calculations about sample means.

🧠 Quick Revision Questions

1. What is the mean of the sampling distribution of the sample mean called, and what is its relationship to the population mean?
2. What is the formula for the standard error of the sample mean when sampling with replacement?
3. When is the finite population correction factor required?
4. State the Central Limit Theorem. Under what condition (sample size) can we assume the sampling distribution of X̄ is normal?
5. If a population is normally distributed, does the sampling distribution of X̄ still require a large sample size (n≥30) to be normal?

📘 Lecture 32 — Sampling Distribution of ( \bar{X}_1 - \bar{X}_2 )

📖 Overview: This lecture continues the discussion of sampling distributions by examining the distribution of the sample mean and sample proportion in practical scenarios. It then introduces the sampling distribution of the difference between two sample means, explaining its properties, mean, standard error, and shape under different conditions. This is crucial for comparing two populations.

🗂️ Topics Covered

The lecture begins by applying the sampling distribution of the sample mean to a real-world problem involving employee salaries. It then introduces the sampling distribution of the sample proportion, defining it, illustrating its properties with an example, and covering its mean, standard error, and the normal approximation condition. Finally, it discusses the sampling distribution of the difference between two sample means, including its mean, standard error, and shape, supported by a detailed example.

📝 Lecture Summary

Application of Sampling Distribution of ( \bar{X} ) (Example)

A construction company has 310 employees with an average salary of Rs.24,000 and a standard deviation of Rs.5,000. A random sample of 100 employees is drawn to check if the average salary exceeds Rs.24,500. Because the sample size (n=100) is greater than 5% of the population (N=310), the finite population correction factor (fpc) is used to compute the standard error of the mean: (\sigma_{\bar{X}} = \frac{\sigma}{\sqrt{n}} \sqrt{\frac{N-n}{N-1}} = \frac{5000}{\sqrt{100}} \sqrt{\frac{310-100}{310-1}} = 412.20).

To find the probability (P(\bar{X} > 24500)), the z-score is computed: (z = \frac{24500-24000}{412.20} = 1.21). Using the standard normal table, (P(\bar{X} > 24500) = 0.5 - 0.3869 = 0.1131). This 11.31% probability suggests the government is likely (89% chance) to find the demand valid.

💡 Why this matters: This demonstrates how the Central Limit Theorem and the concept of sampling distribution are applied in real-world statistical inference, such as hypothesis testing and decision-making.

Sampling Distribution of the Sample Proportion (( \hat{p} ))

When population elements are classified into two categories (e.g., success/failure), the population proportion (p = X/N) and the sample proportion (\hat{p} = x/n), where (x) is the number of successes in the sample. The sampling distribution of (\hat{p}) is the probability distribution of the proportions of successes in all possible random samples of size (n).

🔑 Definition — Sample Proportion: A random variable computed as the number of successes in the sample divided by the sample size.

Example (with verification): A population of six values (1,3,6,8,9,12) is used. All 20 possible samples of size 3 (without replacement) are drawn. The population proportion of even numbers is (p = 3/6 = 0.5). The sampling distribution of (\hat{p}) is constructed.

• Mean: (\mu_{\hat{p}} = \sum \hat{p} f(\hat{p}) = 10/20 = 0.5), verifying (\mu_{\hat{p}} = p).
• Variance: (\sigma^2_{\hat{p}} = \sum \hat{p}^2 f(\hat{p}) - [\sum \hat{p} f(\hat{p})]^2 = 6/20 - (10/20)^2 = 0.05). This matches (\frac{pq}{n} \cdot \frac{N-n}{N-1} = \frac{(0.5)(0.5)}{3} \cdot \frac{3}{5} = 0.05).

📐 Properties of Sampling Distribution of ( \hat{p} ):

• Mean: (\mu_{\hat{p}} = p)
• Standard Error:
  • With replacement: (\sigma_{\hat{p}} = \sqrt{\frac{pq}{n}})
  • Without replacement (finite population): (\sigma_{\hat{p}} = \sqrt{\frac{pq}{n} \cdot \frac{N-n}{N-1}})
• Shape: Approximately normal if both (np \ge 5) and (nq \ge 5). Continuity correction of (\pm \frac{1}{2n}) is applied for normal approximation.

Example – 10% of sugar boxes are underweight (p=0.10). For a sample of n=144, find (P(\hat{p} \ge 0.05)). Applying continuity correction ((\hat{p} \ge 0.05 - \frac{1}{2(144)} = 0.04653)). (\sigma_{\hat{p}} = \sqrt{\frac{(0.1)(0.9)}{144}} = 0.025). (z = \frac{0.04653 - 0.10}{0.025} = -2.14). Thus, (P(Z \ge -2.14) = 0.4838 + 0.5 = 0.9838).

Sampling Distribution of the Difference Between Two Means (( \bar{X}_1 - \bar{X}_2 ))

This distribution is used to compare two populations. It is the probability distribution of the differences between the means of independent samples from two populations.

📌 Example: Draw all possible samples of size (n_1=2) (with replacement) from Population 1: {4,6,8} and size (n_2=2) from Population 2: {1,2,3}. Compute all 81 differences (\bar{x}_1 - \bar{x}_2). The sampling distribution is constructed, yielding mean = 4 and variance = 5/3 ≈ 1.67.

Verification:

• Mean of Population 1: (\mu_1 = 6), Variance: (\sigma^2_1 = 8/3).
• Mean of Population 2: (\mu_2 = 2), Variance: (\sigma^2_2 = 2/3).
• (\mu_{\bar{X}_1 - \bar{X}_2} = 4 = 6 - 2 = \mu_1 - \mu_2).
• (\sigma^2_{\bar{X}_1 - \bar{X}_2} = \frac{\sigma^2_1}{n_1} + \frac{\sigma^2_2}{n_2} = \frac{8/3}{2} + \frac{2/3}{2} = 4/3 + 1/3 = 5/3 \approx 1.67). The computed variance matches.

🔑 Properties of Sampling Distribution of ( \bar{X}_1 - \bar{X}_2 ):

1. Mean: (\mu_{\bar{X}_1 - \bar{X}_2} = \mu_1 - \mu_2)
2. Standard Error: (\sigma_{\bar{X}_1 - \bar{X}_2} = \sqrt{\frac{\sigma^2_1}{n_1} + \frac{\sigma^2_2}{n_2}}) (for infinite populations or sampling with replacement). For sampling without replacement from finite populations, a finite population correction factor is used.
3. Shape:
   • If both populations are normally distributed, the distribution of (\bar{X}_1 - \bar{X}_2) is exactly normal regardless of sample sizes.
   • If populations are non-normal but both sample sizes are large (n ≥ 30), the distribution is approximately normal (Central Limit Theorem). The test statistic (Z = \frac{(\bar{X}_1 - \bar{X}2) - (\mu_1 - \mu_2)}{\sigma{\bar{X}_1 - \bar{X}_2}}) is approximately N(0,1).

💡 Why this matters: This distribution is the foundation for hypothesis tests and confidence intervals comparing the means of two independent groups (e.g., comparing test scores of two classes or effectiveness of two drugs).

⭐ Key Takeaways

1. The finite population correction factor is essential when sampling without replacement from a finite population where sample size exceeds 5% of the population size. It increases the precision of the standard error.
2. The sampling distribution of the sample proportion (\hat{p}) has mean (p) and standard error (\sqrt{pq/n}) (with fpc for finite populations). Its shape is approximately normal when both (np) and (nq) are at least 5, requiring a continuity correction of (\pm 1/(2n)).
3. The sampling distribution of the difference between two means (\bar{X}_1 - \bar{X}_2) has a mean equal to (\mu_1 - \mu_2) and variance equal to (\sigma^2_1/n_1 + \sigma^2_2/n_2).
4. The shape of the distribution of (\bar{X}_1 - \bar{X}_2) is exactly normal if both populations are normal, and approximately normal (via CLT) if both sample sizes are large (≥30), regardless of the population distribution.
5. Any probability calculation involving sample statistics (mean or proportion) follows a three-step process: standardize the statistic using its properties, apply continuity correction (for proportion), and use the standard normal table to find the area/probability.

🧠 Quick Revision Questions

1. When is the finite population correction factor (fpc) used in the sampling distribution of the mean, and how does it affect the standard error?
2. What are the three main properties (mean, variance, shape) for the sampling distribution of the sample proportion (\hat{p})?
3. For what conditions does the sampling distribution of the difference between two sample means ((\bar{X}_1 - \bar{X}_2)) follow an exact normal distribution? When is it approximately normal?
4. In the example of sugar boxes, why was it necessary to apply a continuity correction of (1/(2n))? What would have happened if it was omitted?
5. State the formula for the standard error of (\bar{X}_1 - \bar{X}_2). How does this formula change when sampling without replacement from finite populations?

📘 Lecture 33 — Sampling Distribution of (continued) & Point Estimation

📖 Overview: This lecture continues the discussion of sampling distributions, specifically focusing on the sampling distribution of the difference between two sample means and the sampling distribution of the difference between two sample proportions. It then introduces the fundamental concepts of point estimation, including the distinction between an estimator and an estimate, and examines two desirable qualities of a good point estimator: unbiasedness and consistency.

🗂️ Topics Covered

The lecture covers a practical example applying the sampling distribution of the difference between two means, then introduces the sampling distribution of the difference between two proportions along with its properties and an example. The discussion shifts to point estimation, differentiating between an estimator and an estimate, and explains the concept of standard error. Finally, it details the desirable qualities of a good point estimator, specifically unbiasedness and consistency, with examples and formula derivations.

📝 Lecture Summary

Sampling Distribution of X̅₁ - X̅₂ (continued)

The lecture begins with a real-life application of the sampling distribution of the difference between two sample means, illustrating how to compute probabilities concerning differences in sample means from two independent populations.

🔑 Definition — Sampling Distribution of X̅₁ - X̅₂: The probability distribution of all possible differences between two independent sample means drawn from two populations.

📐 Formula: For large sample sizes (n₁, n₂ ≥ 30), the sampling distribution of X̅₁ - X̅₂ is approximately normal with:

• Mean: μₓ₁₋ₓ₂ = μ₁ - μ₂
• Standard Error: σₓ₁₋ₓ₂ = √(σ₁²/n₁ + σ₂²/n₂)

The standardized variable is: Z = [(X̅₁ - X̅₂) - (μ₁ - μ₂)] / √(σ₁²/n₁ + σ₂²/n₂)

📌 Example: Car batteries from company A have μ₁ = 4.3 years, σ₁ = 0.6 years. Company B batteries have μ₂ = 4.0 years, σ₂ = 0.4 years. Find the probability that a random sample of 49 batteries from A will have a mean life at least 0.5 years more than the mean life of a sample of 36 batteries from B.

Step 1: Compute the mean of the sampling distribution: μₓ₁₋ₓ₂ = 4.3 - 4.0 = 0.3 years

Step 2: Compute the standard error: σₓ₁₋ₓ₂ = √((0.6)²/49 + (0.4)²/36) = √(0.36/49 + 0.16/36) = √(0.00735 + 0.00444) = √0.01179 = 0.1086 years

Step 3: Compute the z-value for X̅₁ - X̅₂ = 0.5: z = (0.5 - 0.3) / 0.1086 = 0.2 / 0.1086 = 1.84

Step 4: Find the probability: P(X̅₁ - X̅₂ ≥ 0.5) = P(Z ≥ 1.84) = 0.5 - P(0 < Z < 1.84) = 0.5 - 0.4671 = 0.0329

The probability is 3.29%. Given the true difference in mean lifetimes is 0.3 years, there is only a 3.3% chance of observing a sample difference of at least 0.5 years. 💡 Why this matters: This demonstrates how to assess whether an observed difference between two sample means is statistically significant or could have occurred by random chance.

Sampling Distribution of the Differences Between Proportions

This section introduces the sampling distribution for comparing proportions from two independent binomial populations.

🔑 Definition — Sampling Distribution of p̂₁ - p̂₂: The probability distribution of the differences between the proportions of success from all possible pairs of independent samples drawn from two binomial populations with proportions p₁ and p₂.

Properties of the Sampling Distribution of p̂₁ - p̂₂:

Property No. 1 (Mean): μₚ̂₁₋ₚ̂₂ = p₁ - p₂

Property No. 2 (Standard Error): σₚ̂₁₋ₚ̂₂ = √(p₁q₁/n₁ + p₂q₂/n₂), where q = 1 - p

📌 Example: It is claimed that 30% of households in Community A and 20% in Community B have at least one teenager. A sample of 100 households from each community yields p̂ₐ = 0.34 and p̂_B = 0.13. What is the probability of observing a difference this large or larger if the claims are true?

Step 1: State the population parameters: pₐ = 0.30, p_B = 0.20 (claims assumed true) nₐ = n_B = 100

Step 2: Compute the mean of the sampling distribution: μₚ̂ₐ₋ₚ̂_B = 0.30 - 0.20 = 0.10

Step 3: Compute the standard error: σₚ̂ₐ₋ₚ̂_B = √((0.30)(0.70)/100 + (0.20)(0.80)/100) = √(0.21/100 + 0.16/100) = √(0.0021 + 0.0016) = √0.0037 = 0.0608

Step 4: Observed difference: p̂ₐ - p̂_B = 0.34 - 0.13 = 0.21

Step 5: Compute z-value: z = (0.21 - 0.10) / 0.0608 = 0.11 / 0.0608 = 1.83

Step 6: Find the probability: From the standard normal table, area between z = 0 and z = 1.83 is 0.4664. P(p̂ₐ - p̂_B ≥ 0.21) = P(Z ≥ 1.83) = 0.5 - 0.4664 = 0.0336

The probability is 3.36%. If the claims are true, the chance of observing a difference as large as or larger than 0.21 is only 3.36%.

Note on Sampling Distributions:

• All discussed sampling distributions assume simple random sampling
• This technique is appropriate when the population is homogeneous

Why "Standard Error"? The term "standard error" is used because any statistic considered as an estimate of a population parameter will have a difference (error) from the true parameter value. The standard deviation of the sampling distribution provides a "standard" value of this error.

Point Estimation

This section introduces the fundamental concepts of estimating population parameters using sample data.

🔑 Definition — Point Estimation: The process of providing a single value calculated from a sample that is likely to be close in magnitude to the unknown population parameter.

Difference Between 'Estimate' and 'Estimator':

• Estimator: The rule, method, or formula used to estimate a parameter (a random variable)
• Estimate: The numerical value obtained by substituting sample observations into the estimator formula

📌 Example: For a random sample X₁, X₂, ..., Xₙ from a population with mean μ:

• Estimator: X̅ = (1/n) Σᵢ₌₁ⁿ Xᵢ
• Estimate: x̅ (the numerical value of X̅)

An estimator is always a statistic which is a function of sample observations and hence is a random variable. In repeated sampling, an estimator has a probability distribution known as its sampling distribution.

Desirable Qualities of a Good Point Estimator

The lecture covers three criteria for evaluating estimators, with two discussed in detail.

1. Unbiasedness

🔑 Definition — Unbiased Estimator: An estimator θ̂ is unbiased if its expected value equals the true value of the population parameter being estimated. That is, E(θ̂) = θ.

If E(θ̂) ≠ θ, the statistic is called a biased estimator.

Examples of Unbiased Estimators:

• Sample mean X̅: E(X̅) = μ (always unbiased for μ, regardless of sample size)
• Sample median X̃: E(X̃) = μ when the population is normally distributed
• Sample proportion p̂: E(p̂) = p (unbiased estimator of population proportion)

Proof for Sample Proportion: Since X (number of successes) follows binomial distribution with E(X) = np: E(p̂) = E(X/n) = (1/n)E(X) = (1/n)(np) = p

Visual Representation: Unbiasedness means the sampling distribution of X̅ is centered at μ. Over-estimates and under-estimates balance each other out in the long run.

Sample Variance — A Special Case: The sample variance S² = (1/n) Σ(Xᵢ - X̅)² is a biased estimator of σ² because E(S²) ≠ σ².

Modified Formula for Unbiased Sample Variance: s² = Σ(xᵢ - x̅)² / (n - 1) Since E(s²) = σ², s² is an unbiased estimator of σ².

Why Unbiasedness Matters: In an actual study, we can never guarantee that our sample statistic will exactly equal the population parameter. Unbiasedness ensures that our estimation procedure (formula) is such that, in repeated sampling, the average value of our statistic equals the parameter.

Bias Direction:

• Positively biased: E(θ̂) > θ
• Negatively biased: E(θ̂) < θ

2. Consistency

🔑 Definition — Consistent Estimator: An estimator θ̂ is consistent if, for any arbitrarily small positive quantity ε: lim (n→∞) P(|θ̂ - θ| ≤ ε) = 1

In other words, the probability that θ̂ is very close to θ approaches unity as the sample size increases.

Key Properties of Consistency:

• It is a large sample property
• A consistent estimator may or may not be unbiased

Examples:

• Sample mean X̅: Consistent estimator of μ (and unbiased)
• Sample proportion p̂: Consistent estimator of p
• Sample median: NOT a consistent estimator of μ when the population has a skewed distribution
• Sample variance S²: Though biased, IS a consistent estimator of σ²

General Rule: A statistic whose standard error decreases with an increase in sample size will be consistent.

⭐ Key Takeaways

1. The sampling distribution of the difference between two sample means (X̅₁ - X̅₂) is approximately normal for large samples, with mean equal to the difference in population means and standard error equal to the square root of the sum of each population variance divided by its respective sample size.
2. The sampling distribution of the difference between two sample proportions (p̂₁ - p̂₂) has mean equal to p₁ - p₂ and standard error equal to √(p₁q₁/n₁ + p₂q₂/n₂), enabling probability calculations for observed differences under assumed population proportions.
3. A point estimator is a random variable (a formula/rule), while an estimate is the numerical value obtained from a specific sample—this distinction is critical in statistical inference.
4. Unbiasedness requires E(θ̂) = θ; the sample mean X̅, sample proportion p̂, and the modified sample variance s² (with n-1 denominator) are unbiased estimators, while S² (with n denominator) is biased.
5. Consistency requires that as sample size increases, the estimator's probability of being close to the parameter approaches 1—this is a large-sample property linked to a decreasing standard error.

🧠 Quick Revision Questions

1. For the battery example, what is the mean and standard error of the sampling distribution of X̅₁ - X̅₂, and what does the computed probability of 0.0329 tell us?
2. What are the two key properties of the sampling distribution of p̂₁ - p̂₂, and how are they used to find the probability of observing a specific sample difference?
3. What is the precise distinction between an "estimator" and an "estimate," and why is an estimator considered a random variable?
4. Explain why the sample variance S² = (1/n)Σ(Xᵢ - X̅)² is a biased estimator of σ², and how the formula is modified to achieve unbiasedness.
5. Define consistency and explain why the sample median is not a consistent estimator of μ when the population distribution is skewed, but the sample mean is consistent.

📘 Lecture 34 — Desirable Qualities of a Good Point Estimator: Efficiency; Methods of Point Estimation: The Method of Moments, The Method of Least Squares, The Method of Maximum Likelihood; Interval Estimation: Confidence Interval for μ

📖 Overview: This lecture completes the discussion of desirable qualities of point estimators by introducing efficiency, then presents three major methods of point estimation: the Method of Moments, Method of Least Squares, and Method of Maximum Likelihood. The lecture concludes by introducing interval estimation, developing the 95% confidence interval for the population mean μ using the normal distribution.

🗂️ Topics Covered

The lecture covers efficiency as a desirable quality of point estimators, followed by three methods of point estimation: the Method of Moments (with examples for uniform and normal distributions), the Method of Least Squares (linked to regression), and the Method of Maximum Likelihood (with examples for Poisson, geometric, Bernoulli, exponential, and normal distributions). The lecture then transitions to interval estimation, explaining the need for confidence intervals and deriving the 95% confidence interval for μ using the standard normal distribution and estimated standard deviation.

📝 Lecture Summary

Desirable Qualities of a Good Point Estimator: Efficiency

An unbiased estimator is defined to be efficient if the variance of its sampling distribution is smaller than that of the sampling distribution of any other unbiased estimator of the same parameter. In other words, suppose that there are two unbiased estimators T₁ and T₂ of the same parameter θ. Then, the estimator T₁ will be said to be more efficient than T₂ if Var(T₁) < Var(T₂).

The relative efficiency of T₁ compared to T₂ (where both are unbiased estimators) is given by the ratio: 🔑 Definition — Relative Efficiency: E = Var(T₂) / Var(T₁) If multiplied by 100, we obtain the relative efficiency in percentage form.

📌 Example: Both the sample mean and the sample median for a population that has a normal distribution are unbiased and consistent estimators of μ, but the variance of the sampling distribution of sample means is smaller than the variance of the sampling distribution of sample medians. Hence, the sample mean is more efficient than the sample median as an estimator of μ.

The Method of Moments

The Method of Moments, due to Karl Pearson (1857-1936), consists of calculating a few moments of the sample values and equating them to the corresponding moments of a population, thus getting as many equations as needed to solve for the unknown parameters.

Let X₁, X₂, ..., Xₙ be a random sample of size n from a population. Then the rth sample moment about zero is: 🔑 Definition — Sample Moment: m'ᵣ = (∑Xᵢʳ) / n, r = 1, 2, ... The corresponding rth population moment is μ'ᵣ. We match these moments and get as many equations as needed to solve for the unknown parameters.

📌 Example 1: Let X be uniformly distributed on the interval (0, θ). Find an estimator of θ by the method of moments. Solution: The probability density function is f(x) = 1/θ, 0 ≤ x ≤ θ. The first sample moment about zero is m'₁ = (∑Xᵢ)/n = X̄. The first population moment about zero is μ'₁ = ∫₀^θ x·(1/θ) dx = [x²/(2θ)]₀^θ = θ/2. Matching: X̄ = θ/2, so θ̂ = 2X̄. Hence, the moment estimator of θ is twice the sample mean.

📌 Example 2: Let X₁, X₂, ..., Xₙ be a random sample of size n from a normal population with parameters μ and σ². Find these parameters by the method of moments. Solution: The first two sample moments: m'₁ = (1/n)∑Xᵢ = X̄ and m'₂ = (1/n)∑Xᵢ². The corresponding moments of a normal distribution: μ'₁ = μ and μ'₂ = σ² + μ². Matching: μ̂ = (1/n)∑Xᵢ = X̄, and σ̂² = (∑Xᵢ²)/n - X̄² = (1/n)∑(Xᵢ - X̄)² = S².

💡 Why this matters: A shortcoming of this method is that the moment estimators are, in general, inefficient.

The Method of Least Squares

The Method of Least Squares, due to Gauss (1777-1855) and Markov (1856-1922), is based on the theory of linear estimation. An estimator found by minimizing the sum of squared deviations of the sample values from some function hypothesized as a fit for the data is called the least squares estimator.

This method was discussed in connection with regression analysis (Lecture No. 15). When fitting a straight line y = a + bx to real data, 'a' and 'b' are determined by minimizing the sum of squared deviations between the fitted line and the data-points. The y-intercept and slope 'a' and 'b' are least-square estimates of the y-intercept and slope of the true line for the entire population.

Method of Maximum Likelihood

The Method of Maximum Likelihood, regarded as the most important method of estimation, was introduced in 1922 by Sir Ronald A. Fisher (1890-1962).

Rationale: "To consider every possible value that the parameter might have, and for each value, compute the probability that the given sample would have occurred if that were the true value of the parameter. That value of the parameter for which the probability of a given sample is greatest, is chosen as an estimate." An estimate obtained by this method is called the maximum likelihood estimate (MLE). This method is applicable to both discrete and continuous random variables.

EXAMPLES OF MLE's IN CASE OF DISCRETE DISTRIBUTIONS:

• Poisson distribution (P(X=x) = e^(-μ)μ^x / x!, x=0,1,2,...): MLE of μ is X̄ (sample mean)
• Geometric distribution: MLE of p is 1/X̄ (reciprocal of the mean)
• Bernoulli distribution (P(X=x) = p^x q^(1-x), x=0,1): MLE of p is X̄ (sample mean)

EXAMPLES OF MLE's IN CASE OF CONTINUOUS DISTRIBUTIONS:

• Exponential distribution (f(x) = θe^(-θx), x>0, θ>0): MLE of θ is 1/X̄ (reciprocal of sample mean)
• Normal distribution (parameters μ and σ²): Joint ML estimators of μ and σ² are X̄ (sample mean) and S² (sample variance, which is not an unbiased estimator of σ²)

💡 Why this matters: Maximum Likelihood Estimators are consistent and efficient but not necessarily unbiased (as S² is not unbiased).

📌 Example: Human weight is approximately normally distributed. A random sample of 15 adult males yields weights (in pounds): 131.5, 136.9, 133.8, 130.1, 133.9, 135.2, 129.6, 134.4, 130.5, 134.2, 131.6, 136.7, 135.8, 134.5, 132.7. Find the MLEs for μ and σ². Solution: X̄ = 133.43 and S² = 5.10. These are the joint Maximum Likelihood Estimates of the mean and variance.

Interval Estimation: Confidence Interval for μ

Point estimation gives a single quantity computed from the sample as an estimate of a population parameter. The limitation is that we have no way of ascertaining how close our point estimate is to the true value.

Interval estimation is a procedure of estimating the unknown parameter which specifies a range of values within which the parameter is expected to lie. A confidence interval is an interval computed from the sample observations with a statement of how confident we are that the interval does contain the population parameter.

The lecture uses the Ministry of Transport test example (faults per car, uniformly distributed 0-4) to show that as sample size increases, the sampling distribution of X̄ becomes approximately normal.

For large samples from an infinite population, X̄ is normally distributed with mean μ and variance σ²/n, i.e., X̄ ~ N(μ, σ²/n). The standardized version Z = (X̄ - μ) / (σ/√n) is N(0,1).

For the standard normal distribution: 🔑 Probability Statement: P(-1.96 < Z < 1.96) = 0.95

This leads to: P(-1.96 ≤ (X̄ - μ) / (σ/√n) ≤ 1.96) = 0.95

Rearranging: P(X̄ - 1.96(σ/√n) ≤ μ ≤ X̄ + 1.96(σ/√n)) = 0.95

The 95% confidence interval for μ is: 📐 Formula: X̄ ± 1.96(σ/√n)

Since the population standard deviation σ is usually unknown, we estimate it. The quantity s² = ∑(Xᵢ - X̄)² / (n-1) is an unbiased estimator of σ².

When σ is unknown, the 95% Confidence Interval for μ is: 📐 Formula: X̄ ± 1.96(s/√n) where s/√n is the standard error of the sample mean.

The points X̄ - 1.96(s/√n) and X̄ + 1.96(s/√n) are called the lower and upper limits of the 95% confidence interval.

⭐ Key Takeaways

You must remember that efficiency compares unbiased estimators by their sampling variance — the one with smaller variance is more efficient, and the sample mean is more efficient than the sample median for normal populations. The three methods of point estimation each have distinct approaches: the Method of Moments equates sample moments to population moments (giving θ̂ = 2X̄ for uniform(0,θ) and X̄, S² for normal), the Method of Least Squares minimizes sum of squared deviations (used in regression), and the Method of Maximum Likelihood maximizes the likelihood function (giving MLEs that are consistent and efficient but not necessarily unbiased). For the normal distribution, the MLEs are X̄ and S². Finally, the 95% confidence interval for μ is X̄ ± 1.96(σ/√n) when σ is known, and X̄ ± 1.96(s/√n) when σ is estimated by s, with the interpretation that we are 95% confident the interval contains the true population mean.

🧠 Quick Revision Questions

1. What does it mean for one unbiased estimator to be more efficient than another?
2. For a uniform distribution on (0, θ), what is the method of moments estimator of θ, and how is it derived?
3. What is the rationale behind the method of maximum likelihood?
4. What are the maximum likelihood estimators of μ and σ² for a normal distribution, and are they unbiased?
5. Write the formula for the 95% confidence interval for μ when the population standard deviation σ is unknown. What quantity is used to estimate σ?

📘 Lecture 35 — Confidence Interval for μ (continued). Confidence Interval for μ₁-μ₂.

📖 Overview: This lecture extends the construction of confidence intervals for a single population mean (μ) with detailed examples and important interpretations regarding the meaning of confidence levels. It then introduces the concept of confidence intervals for the difference between two population means (μ₁-μ₂), providing formulas and worked examples for comparing two distinct groups.

🗂️ Topics Covered

The lecture continues the discussion on confidence intervals for the mean of a population (μ), presenting examples with large samples to construct 95%, 90%, and 99% intervals. It covers the crucial interpretation of confidence in repeated sampling, defines a confidence interval as a point estimate plus/minus a number of standard errors, and discusses the trade-offs between confidence level and interval width. The second half introduces the formula and examples for the confidence interval for the difference between two population means (μ₁-μ₂) using large, independent samples.

📝 Lecture Summary

Confidence Interval for μ (continued).

This section provides worked examples of constructing confidence intervals for the population mean μ. It emphasizes the use of the sample standard deviation (s) as an estimate for the population standard deviation (σ) when the sample size is large (n ≥ 30). The lecture also details the correct interpretation of a 95% confidence interval, clarifying that the probability statement applies to the procedure in repeated sampling, not to a single, specific interval.

🔑 Definition — Point Estimate: The sample mean (X̄) is the best single-number estimate of the population mean μ. 📐 Formula: For large n (n ≥ 30), the confidence interval for μ is X̄ ± z_(α/2) * (s / √n), where z_(α/2) is the critical value from the standard normal distribution. 📌 Example (95% C.I.): A sample of 500 employees has X̄ = 5.38 days and s = 3.53 days. The 95% confidence interval is 5.38 ± 1.96*(3.53/√500) = 5.38 ± 0.31, resulting in an interval of (5.07, 5.69) days. 💡 Why this matters: The statement "the mean lies between 5.07 and 5.69 days with 95% confidence" means that if we were to take many such samples and compute intervals for each, approximately 95% of those intervals would contain the true population mean μ.

📌 Example (90% C.I.): A sample of 64 days on a bridge has X̄ = 5410 and s = 680. The 90% confidence interval is 5410 ± 1.645*(680/√64) = 5410 ± 139.8, resulting in an interval of (5270.2, 5549.8) or (5270, 5550). 📌 Example (99% C.I.): A sample of 110 rental cars has X̄ = 85.5 miles and s = 19.3 miles. The 99% confidence interval is 85.5 ± 2.575*(19.3/√110) = 85.5 ± 4.7, resulting in an interval of (80.8, 90.2) miles per day.

An Important Way of Interpreting a Confidence Interval

This section reframes the confidence interval formula as a point estimate plus/minus a multiple of the standard error of the estimate. It explains that the "multiple" (z_(α/2)) determines the level of confidence, and that a narrower interval with high confidence is preferable. The lecture discusses the two key factors that affect the width of the interval: the sample size (n) and the level of confidence (1-α).

📐 Formula (Reframed): The confidence interval for μ can be defined as X̄ ± z_(α/2) * σ_X̄, where σ_X̄ (sigma sub X-bar) is the standard error of the mean, equal to σ/√n. 📌 Key Concept: A narrow confidence interval is better. It is achieved by making z_(α/2) * σ_X̄ small. This happens when n is large (making σ_X̄ small) or when the confidence level 1-α is low (making z_(α/2) small). In practice, a compromise must be struck between the desired level of confidence and the acceptable width of the interval.

Confidence Interval for the Difference Between the Means of Two Populations

This section introduces the formula for constructing a confidence interval for the difference between two population means (μ₁-μ₂) using large samples drawn independently from two populations.

🔑 Definition — Difference of Means (μ₁-μ₂): The parameter of interest is the real difference between the average values of two distinct populations. 📐 Formula: For large, independent samples, the confidence interval for μ₁ - μ₂ is (X̄₁ - X̄₂) ± z_(α/2) * √( (s₁² / n₁) + (s₂² / n₂) ). 📌 Example 1 (90% C.I. for Income Difference): Factory A (n=160, X̄=12.80, s²=64) and Factory B (n=220, X̄=11.25, s²=47). The 90% C.I. is (12.80 - 11.25) ± 1.645 * √(64/160 + 47/220) = 1.55 ± 1.28, resulting in an interval of (0.27, 2.83). We are 90% confident the true difference in weekly incomes is between Rs. 0.27 and Rs. 2.83. 📌 Example 2 (98% C.I. for Coupon Savings Difference): Middle-income shoppers (n₁=60, X̄₁=$5.84, s₁=$1.41) and Low-income shoppers (n₂=80, X̄₂=$2.67, s₂=$0.54). The 98% C.I. is (5.84 - 2.67) ± 2.33 * √((1.41²/60) + (0.54²/80)) = 3.17 ± 0.45, resulting in an interval of (2.72, 3.62). The data strongly indicates a real difference as the interval does not contain zero.

⭐ Key Takeaways

A confidence interval for a mean (μ) has the form point estimate ± margin of error, where the margin of error is the critical value (z) times the standard error. A 95% confidence interval means that in repeated sampling, 95% of such intervals will contain the true population mean; it does not mean there is a 95% probability for any single computed interval. The width of a confidence interval is inversely related to sample size and directly related to the desired level of confidence. For comparing two population means (μ₁-μ₂) with large, independent samples, the same logic applies, and the interval formula uses the difference of sample means and the combined standard error. If the confidence interval for a difference does not include zero, it provides evidence that a real difference between the two population means exists.

🧠 Quick Revision Questions

1. What is the formula for a (1-α)100% confidence interval for a population mean (μ) when the sample size is large (n ≥ 30)?
2. Explain the correct interpretation of a "95% confidence interval" for a population mean.
3. List the two primary factors that affect the width of a confidence interval and state how changing each factor affects the width.
4. What is the formula for constructing a confidence interval for the difference between two population means (μ₁-μ₂) when using large, independent samples?
5. In Example 2 for the difference of means (coupon savings), the 98% confidence interval was ($2.72, $3.62). What can you conclude about the difference between the means of the two populations?

📘 Lecture 36 — Large Sample Confidence Intervals for p and p1-p2, Determination of Sample Size (with reference to Interval Estimation), Hypothesis-Testing (An Introduction)

📖 Overview: This lecture extends the concept of confidence intervals to population proportions and the difference between two proportions. It then addresses the critical practical question of determining the required sample size for estimation. Finally, it introduces the fundamental concepts of hypothesis-testing, which, alongside estimation, forms the core of statistical inference.

🗂️ Topics Covered

The lecture begins with constructing and interpreting large-sample confidence intervals for a single population proportion (p) and for the difference between two population proportions (p1-p2), illustrated with detailed examples. It then explains how to calculate the necessary sample size for estimating a population mean and a population proportion with a desired precision and confidence level. The lecture concludes by defining what a statistical hypothesis is and introducing the concepts of null and alternative hypotheses, Type-I and Type-II errors, and the level of significance.

📝 Lecture Summary

Large Sample Confidence Intervals for p and p1-p2

The first major topic is constructing confidence intervals for a population proportion, p, which represents the proportion of successes in a binomial population. For a large sample, the confidence interval for p is given by p̂ ± z(α/2) √[p̂(1-p̂)/n], where p̂ is the sample proportion of successes, n is the sample size, and z(α/2) is the critical value from the standard normal distribution (1.96 for 95% confidence, 2.58 for 99% confidence). A sample is considered "large enough" if both np̂ and n(1-p̂) are greater than or equal to 5.

🔑 Definition — Large Sample Confidence Interval for p: A range of values, calculated from sample data, that is likely to contain the true population proportion p with a specified level of confidence.

📐 Formula: p̂ ± z(α/2) * √[p̂(1-p̂)/n] → This formula calculates the margin of error around the point estimate (p̂) to create an interval that is expected to contain the true population proportion p a certain percentage of the time.

📌 Example (EXAMPLE-1):

• Context: A survey of 634 teenagers with three or more juvenile court appearances found 291 were orphans.
• Goal: Estimate the proportion of all such teenagers who are orphans with 99% confidence.
• Steps:
  1. Identify variables: n = 634, number of successes = 291.
  2. Calculate sample proportion (p̂): p̂ = 291/634 = 0.459. Then q̂ = 1 - 0.459 = 0.541.
  3. Find critical value: For 99% confidence, z(α/2) = 2.58.
  4. Calculate standard error: √[p̂*q̂/n] = √[(0.459 * 0.541) / 634].
  5. Compute margin of error: 2.58 * standard error = 0.051.
  6. Construct interval: 0.459 ± 0.051 = (0.408, 0.510).
• Conclusion: We are 99% confident that the true proportion of these teenagers who are orphans is between 40.8% and 51.0%.

📌 Example (EXAMPLE-2):

• Context: A poll of 500 voters found 290 favor Mr. Scott.
• Goal: Develop a 95% confidence interval for the proportion of all voters who will vote for Mr. Scott and determine if he will win.
• Steps:
  1. Identify variables: n = 500, number of successes = 290. p̂ = 290/500 = 0.58.
  2. Find critical value: For 95% confidence, z(α/2) = 1.96.
  3. Calculate standard error: √[0.58*(1-0.58)/500] = √[(0.58*0.42)/500].
  4. Compute margin of error: 1.96 * standard error = 0.043.
  5. Construct interval: 0.58 ± 0.043 = (0.537, 0.623).
• Conclusion: Since the entire confidence interval (0.537 to 0.623) is above 0.50, we conclude that Mr. Scott will likely win the election.

📌 Example (EXAMPLE-3):

• Context: A survey of 210 chief executives found that 51% had a management-succession plan.
• Goal: Compute a 92% confidence interval to estimate the true proportion of all fast-growing small companies that have such a plan.
• Steps:
  1. Identify variables: n = 210, p̂ = 0.51, q̂ = 0.49.
  2. Find critical value: For 92% confidence, z(α/2) = 1.75.
  3. Calculate standard error: √[(0.51*0.49)/210].
  4. Compute margin of error: 1.75 * standard error = 0.06.
  5. Construct interval: 0.51 ± 0.06 = (0.45, 0.57).
• Conclusion: We are 92% confident that the true proportion is between 0.45 and 0.57.

The lecture then shifts to the confidence interval for the difference in two population proportions, (p1-p2). For large independent samples from two binomial populations, the confidence interval is (p̂1 – p̂2) ± z(α/2) √[ p̂1(1-p̂1)/n1 + p̂2(1-p̂2)/n2 ].

📐 Formula: (p̂1 – p̂2) ± z(α/2) * √[ p̂1(1-p̂1)/n1 + p̂2(1-p̂2)/n2 ] → This formula estimates the difference between two population proportions by creating an interval around the observed difference in sample proportions.

📌 Example:

• Context: Among college students, 300 of 400 resident students and 200 of 300 non-resident students approved a course of action.
• Goal: Estimate the difference in proportions (resident minus non-resident) and compute a 90% confidence interval.
• Steps:
  1. Identify variables: n1=400, successes1=300, p̂1=0.75; n2=300, successes2=200, p̂2=0.67.
  2. Find critical value: For 90% confidence, z(α/2) = 1.645.
  3. Calculate standard error: √[(0.750.25)/400 + (0.670.33)/300] = √[0.00046875 + 0.000737] = √0.00120575 ≈ 0.0347.
  4. Compute margin of error: 1.645 * 0.0347 = 0.057.
  5. Construct interval: (0.75 – 0.67) ± 0.057 = 0.08 ± 0.057 = (0.023, 0.137).
• Conclusion: We are 90% confident that the true difference in proportions is between 2.3% and 13.7%.

Determination of Sample Size (with reference to Interval Estimation)

This section addresses the critical question: "How large should the sample be?" The goal is to find a sample size that provides a desired level of precision (error bound, e) with a desired level of confidence.

For estimating a population mean (μ), the required sample size is derived from the error bound formula for a confidence interval. The maximum allowable error, e, is given by e = z(α/2) * σ/√n. Solving for n yields the formula: n = (z(α/2) * σ / e)². Since the population standard deviation σ is usually unknown, an estimate is used, often from a pilot study or past data. The computed value for n is always rounded up to the next integer.

📐 Formula: n = (z(α/2) * σ / e)² → This formula calculates the minimum sample size needed to estimate a population mean with a specified margin of error (e) and confidence level (determined by z(α/2)).

📌 Example (for population mean):

• Context: Estimate the mean of a population with 95% confidence so that sample mean differs from the true mean by no more than 25% of the standard deviation.
• Steps:
  1. Identify variables: e = (25%) of σ = σ/4, z(α/2) for 95% confidence = 1.96.
  2. Apply formula: n = (1.96 * σ / (σ/4))² = (1.96 * 4)² = 61.4656.
• Conclusion: The required sample size is the next highest integer, 62.

For estimating a population proportion (p), the error bound is e = z(α/2) * √[p̂*q̂/n] . Solving for n yields: n = [(z(α/2))² * p̂ * q̂] / e². The values for p̂ and q̂ are unknown before sampling, so a pilot sample estimate is used. If no estimate is available, the most conservative value, p̂=0.5, is used as it maximizes the product p̂*q̂.

📐 Formula: n = [(z(α/2))² * p̂ * q̂] / e² → This formula calculates the minimum sample size needed to estimate a population proportion with a specified margin of error and confidence level.

📌 Example (for population proportion):

• Context: In a pilot sample of 75 axle shafts, 12 had a rough surface. We want to be 95% confident that the error in estimating p is less than 0.05.
• Steps:
  1. Identify variables: p̂ = 12/75 = 0.16, q̂ = 1-0.16 = 0.84, e = 0.05, z(α/2) for 95% = 1.96.
  2. Apply formula: n = (1.96)² / (0.05)² * (0.16 * 0.84) = (3.8416 / 0.0025) * 0.1344 = 1536.64 * 0.1344 = 206.52.
• Conclusion: The required sample size is 207.

Hypothesis-Testing (An Introduction)

This section introduces hypothesis-testing, the second major branch of statistical inference. A statistical hypothesis is a statement or assumption about a population parameter that may or may not be true. The goal of hypothesis testing is to decide, based on sample data, whether to accept or reject such a hypothesis. It is crucial to understand that "accepting" a hypothesis means there is insufficient evidence to reject it, not that it is proven true. "Rejecting" a hypothesis means declaring it false.

🔑 Definition — Null Hypothesis (H0): The hypothesis to be tested for possible rejection or nullification. It is a precise statement, often about no change, no effect, or a specific numerical value. For example, H0: μ = 62 inches.

🔑 Definition — Alternative Hypothesis (H1 or HA): The hypothesis that is accepted when the null hypothesis is rejected. It represents the research claim or suspected difference. For example, H1: μ ≠ 62 inches or H1: μ < 62 inches.

💡 Why this matters: In any hypothesis test, H0 and H1 are mutually exclusive and exhaustive statements about the population parameter. The sample data is used to decide which one is more plausible.

The decision to reject or accept H0 involves a risk of error.

🔑 Definition — Type-I Error: The error of rejecting a null hypothesis when it is actually true. It is also called an error of the first kind.

🔑 Definition — Type-II Error: The error of accepting (not rejecting) a null hypothesis when it is actually false. It is also called an error of the second kind.

🔑 Definition — Level of Significance (α): The probability of committing a Type-I error, P(reject H0 | H0 is true). The probability of a Type-II error is denoted by β.

📌 Example: A judge presumes a defendant is innocent (H0: Innocent).

• Type-I Error: The jury rejects H0 and convicts an innocent person.
• Type-II Error: The jury accepts H0 and acquits a guilty person.

⭐ Key Takeaways

The key learning from this lecture is the ability to construct confidence intervals for proportions and understand the planning needed before data collection. You must be able to apply the correct formula for a single proportion p or the difference between two proportions (p1-p2), and correctly interpret the resulting interval. Equally critical is understanding how sample size determination works, using the formulas n = (zσ/e)² for a mean and n = (z² p̂ q̂)/e² for a proportion, and always rounding up. Finally, you should be able to define the null and alternative hypotheses, and distinguish between Type-I (rejecting a true H0) and Type-II (failing to reject a false H0) errors, as these concepts form the foundation for all hypothesis testing.

🧠 Quick Revision Questions

1. What are the two conditions that must be met for a sample to be considered "large enough" to construct a confidence interval for a population proportion (p)?
2. A researcher wants to estimate the difference in approval rates for a policy between two cities, with a 90% confidence. What is the value of z(α/2) they should use?
3. Explain the difference between a Type-I error and a Type-II error in the context of testing a new drug where H0 is that the drug is ineffective.
4. A researcher needs to estimate a population mean with a margin of error of 2 units and is 99% confident. If a known standard deviation is 8, what sample size is required?
5. You want to estimate the proportion of students who own a car with a 95% confidence and a 3% margin of error. If no pilot study is available, what conservative value of p̂ should you use, and why?

📘 Lecture 37 — Hypothesis-Testing (continuation of basic concepts) & Hypothesis-Testing regarding μ (based on Z-statistic)

📖 Overview: This lecture continues the discussion of hypothesis-testing by introducing the test-statistic, acceptance and rejection regions, and the critical region. It explains the crucial distinction between one-tailed and two-tailed tests and provides detailed, step-by-step examples of hypothesis testing for the population mean using the Z-statistic.

🗂️ Topics Covered

This lecture covers the concept of a test-statistic and its sampling distribution, the division of test-statistic values into acceptance and rejection regions defined by critical values, the relationship between the level of significance (α) and the critical region, and the detailed methodology of one-tailed and two-tailed tests. It concludes with two comprehensive solved examples: a two-tailed test regarding a change in weekly desk production and a one-tailed test regarding day-care costs.

📝 Lecture Summary

Basic Concepts of Hypothesis-Testing (Continuation)

TEST-STATISTIC A test-statistic is a statistic (a function of the sample data not containing any parameters) that provides a basis for testing a null hypothesis. Every test-statistic has a probability distribution (its sampling distribution) which gives the probability that the test-statistic will assume a value greater than or equal to a specified value (or less than or equal to a specified value) when the null hypothesis is true.

ACCEPTANCE AND REJECTION REGIONS All possible values a test-statistic can assume are divided into two mutually exclusive groups. One group consists of values consistent with the null hypothesis, called the acceptance region. The other group leads to the rejection of the null hypothesis and is called the rejection region or critical region. The value(s) separating the critical region from the acceptance region are called the critical value(s). The critical value is decided by the experimenter, with the most frequently used significance levels (α) being 0.05 (5%) and 0.01 (1%). A significance level of α = 5% means there are about 5 chances in 100 of incorrectly rejecting a true null hypothesis (a Type-I error).

RELATIONSHIP BETWEEN THE LEVEL OF SIGNIFICANCE AND THE CRITICAL REGION The level of significance (α) is the basis for determining the critical region of the test. For example, testing H0: μ = 45 against H1: μ ≠ 45 with a 5% significance level and a standard normal Z test-statistic, the critical values are Z = ±1.96. This means the total area of the rejection region (α) is 0.05, equally divided as 0.025 in each tail of the distribution. 💡 Why this matters: The significance level is a direct input for finding the critical values, which define the decision rule for the hypothesis test.

• 🔑 Definition — Test-Statistic: A function of the sample data used to decide whether to reject the null hypothesis.
• 🔑 Definition — Critical Region (Rejection Region): The set of values for the test-statistic that leads to the rejection of H0.
• 🔑 Definition — Critical Value(s): The value(s) that separate the critical region from the acceptance region.
• 📐 Formula — Critical Value for Two-Tailed Test (α=0.05): Z = ±1.96 → The test-statistic must be less than -1.96 or greater than +1.96 to reject H0.

ONE-TAILED AND TWO-TAILED TESTS

One-Tailed (or One-Sided) Test A test where the entire rejection region lies in only one tail of the sampling distribution. This is used when the alternative hypothesis (H1) states a direction: H1: θ > θ0 (right-tailed) or H1: θ < θ0 (left-tailed). For example, for a population mean with a large sample, H1: μ > μ0 uses a right-tailed test.

• For H1: μ < μ0: Reject H0 if the computed Z < –zα.
• For H1: μ > μ0: Reject H0 if the computed Z > zα.

Two-Tailed (or Two-Sided) Test A test where the rejection region is divided equally between the two tails of the sampling distribution. This is used when the alternative hypothesis does not state a direction, e.g., H1: μ ≠ μ0 (meaning H1: μ < μ0 or μ > μ0).

• For H1: μ ≠ μ0: Reject H0 if the computed Z < –zα/2 or if the computed Z > zα/2.

The location of the critical region is determined only after H1 has been stated. One-tailed and two-tailed tests differ only in the location of the critical region, not in its total size (α).

EXAMPLE (Two-Tailed Test)

A steel company’s weekly production of Model A desks has a historical mean of μ=200 and standard deviation σ=16. The vice president wants to know if the mean production last year (n=50, X̄=203.5) differs from 200 at the 0.05 significance level.

SOLUTION:

Step-1: Formulation of Hypotheses H0: μ = 200 (The population mean is 200) H1: μ ≠ 200 (The mean is different from 200) (Note: This is a two-tailed test as no direction is stated.)

Step-2: Level of Significance α = 0.05

Step-3: Test Statistic For a large sample mean: Z = (X̄ – μ) / (σ / √n)

Step-4: Calculations n = 50, X̄ = 203.5, σ = 16 Computed Z = (203.5 – 200) / (16 / √50) = 3.5 / (16 / 7.071) = 3.5 / 2.263 = 1.55

Step-5: Critical Region For a two-tailed test at α=0.05, the critical values are +1.96 and -1.96. The decision rule is: Reject H0 if the computed Z is not between -1.96 and +1.96.

Step-6: Conclusion The computed Z (1.55) lies between -1.96 and +1.96, so H0 is not rejected. The sample evidence does not show that the production rate has changed from 200 per week. The difference of 3.5 units is attributed to chance.

• 📌 Example: The computed z-value of 1.55 was less than the critical value of 1.96, leading to a failure to reject the null hypothesis.

EXAMPLE (One-Tailed Test)

A random sample of 100 workers shows a mean day-care cost of Rs.2650 with a standard deviation of Rs.500. Verify the department’s claim that the mean exceeds Rs.2500 at the 0.05 level.

SOLUTION:

Step-1: Formulation of Hypotheses The claim (mean exceeds Rs.2500) is H1. The negation (mean ≤ Rs.2500) is H0, which contains the equal sign. H0: μ ≤ 2500 H1: μ > 2500 (exceeds 2500)

Step-2: Level of Significance α = 0.05

Step-3: Test Statistic The test-statistic, under H0, is Z = (X̄ – μ0) / (S / √n), which is approximately normal for a large sample (n=100).

Step-4: Critical Region For this right-tailed test, the rejection region is Z > Z0.05 = 1.645.

Step-5: Calculations n = 100, X̄ = 2650, S = 500, μ0 = 2500 Computed Z = (2650 – 2500) / (500 / √100) = 150 / (500 / 10) = 150 / 50 = 3

Step-6: Conclusion The computed Z (3) is greater than the critical value (1.645), so it falls in the rejection region. Therefore, H0 is rejected, and the department’s claim is supported by the sample evidence.

An Interesting and Important Point: For α = 0.01, Zα = 2.33. Since the computed Z (3) is even greater than 2.33, the computed value of X̄ is highly significant. Even with only a 1% chance of being wrong, the department’s claim was correct.

• 📌 Example: The computed z-value of 3 was greater than the critical value of 1.645, leading to the rejection of the null hypothesis.

⭐ Key Takeaways

The most critical points from this lecture are: 1) The test-statistic and its sampling distribution are the mathematical basis for hypothesis testing. 2) The level of significance (α) directly determines the critical value(s) which separate the acceptance and rejection regions. 3) The form of the alternative hypothesis (H1) dictates whether a one-tailed or two-tailed test is used, which changes the location of the critical region. 4) The six-step procedure for hypothesis testing provides a structured framework for making data-driven decisions. 5) A computed test-statistic value can be significant at one level (α=0.05) and highly significant at a stricter level (α=0.01), providing stronger evidence against the null hypothesis.

🧠 Quick Revision Questions

1. What are the six steps in the formal procedure for hypothesis testing presented in this lecture?
2. For a two-tailed test at the 0.01 significance level, what would be the critical Z-values for the test-statistic?
3. How does the location of the critical (rejection) region differ between a one-tailed test (where H1: μ > μ0) and a two-tailed test (where H1: μ ≠ μ0)?
4. In the first example, the computed Z-statistic was 1.55. Was the null hypothesis rejected or not, and what was the conclusion regarding the change in weekly production?
5. In the second example explained why the computed Z-statistic of 3 is considered "highly significant" at the 1% level?

📘 Lecture 38 — Hypothesis-Testing regarding μ1 - μ2 (based on Z-statistic) and Hypothesis Testing regarding p (based on Z-statistic)

📖 Overview: This lecture extends hypothesis testing to comparing two population means using the Z-statistic for large samples, and then moves to testing hypotheses about a population proportion p. It reinforces the six-step hypothesis testing procedure through detailed examples involving wage comparisons between professions and testing proportions of orphans in juvenile court cases.

🗂️ Topics Covered

The lecture reviews the general six-step hypothesis testing procedure, then demonstrates testing the equality of two population means (μ1 - μ2 = 0) using the Z-statistic when sample sizes are large, including handling unknown variances by substituting sample variances. It then addresses hypothesis testing regarding a population proportion p, incorporating the continuity correction for large samples.

📝 Lecture Summary

Hypothesis-Testing regarding μ1 - μ2 (based on Z-statistic)

The lecture begins by restating the general six-step hypothesis testing procedure: (1) formulate H0 and H1, (2) decide significance level α, (3) choose test statistic, (4) determine critical/rejection region, (5) compute test statistic from sample, (6) draw conclusion. We always begin by assuming the null hypothesis is true.

An important note clarifies the relationship between the unbiased estimator s² and the biased estimator S². We know that s² = Σ(x - x̄)²/(n-1) and S² = Σ(x - x̄)²/n. Since (n-1)s² = nS², we get S² = [(n-1)/n]s² = (1 - 1/n)s². As n → ∞, 1/n → 0, so for large n, S² ≈ s². 💡 Why this matters: In large samples, when population variance σ² is unknown, we can replace σ² with S².

When testing the equality of two population means with large samples, the test statistic is: Z = [(X̄₁ - X̄₂) - (μ₁ - μ₂)] / √(σ₁²/n₁ + σ₂²/n₂). When σ₁² and σ₂² are unknown but sample sizes are large (>30), we replace them with S₁² and S₂².

🔑 Definition — Two-sample Z-test for means: A hypothesis test comparing two population means using the standard normal distribution, valid when both sample sizes are large (n₁, n₂ ≥ 30).

📐 Formula: Z = (X̄₁ - X̄₂ - D₀) / √(S₁²/n₁ + S₂²/n₂) → The difference between sample means (minus the hypothesized difference D₀) divided by the standard error of the difference.

📌 Example (Computer Analysts vs. Nurses): A survey 5 years ago showed equal wages. This year, samples of 32 computer analysts and 34 registered nurses are taken to test if wages differ at 2% significance.

• Step 1: H₀: μ₁ - μ₂ = 0 (equal wages), H₁: μ₁ - μ₂ ≠ 0 (two-tailed)
• Step 2: α = 0.02
• Step 3: Z = (X̄₁ - X̄₂ - 0) / √(S₁²/n₁ + S₂²/n₂)
• Step 4: n₁ = 32, X̄₁ = $23.14, S₁² = 1.854; n₂ = 34, X̄₂ = $21.99, S₂² = 1.845. Computed Z = (23.14 - 21.99) / √(1.854/32 + 1.845/34) = 1.15 / 0.335 = 3.43
• Step 5: Critical region: |Z| > 2.33 (since α/2 = 0.01, Z₀.₀₁ = ±2.33)
• Step 6: Since 3.43 > 2.33, reject H₀. Conclusion: There is a significant difference between average hourly wages, with computer analysts earning more.

📌 Example (Factory Workers' Income): Workers of Factory B believe Factory A workers earn more. Samples: Factory A (n₁=160, x̄₁=12.80, s₁²=64), Factory B (n₂=220, x̄₂=11.25, s₂²=47). Test at 5% significance.

• Step 1: H₀: μ₁ - μ₂ ≤ 0, H₁: μ₁ - μ₂ > 0 (right-tailed)
• Step 2: α = 0.05
• Steps 3&4: Z = (12.80 - 11.25 - 0) / √(64/160 + 47/220) = 1.55 / √(0.40 + 0.214) = 1.55 / 0.78 = 1.99
• Step 5: Critical region: Z > 1.645
• Step 6: Since 1.99 > 1.645, reject H₀. The evidence supports the belief that Factory A workers earn more.

Hypothesis Testing regarding p (based on Z-statistic)

This section covers testing a hypothesis about the population proportion p (the proportion of successes). For large samples, the test uses a Z-statistic with a continuity correction of ±½.

🔑 Definition — Continuity correction: An adjustment of ±0.5 applied when using a continuous distribution (normal) to approximate a discrete distribution (binomial), improving accuracy.

📐 Formula: Z = [X ± ½ - np₀] / √[np₀(1-p₀)] → The observed count X (adjusted by ±½) minus the expected count under H₀, divided by the standard deviation of the binomial distribution. Use X - ½ when X > np₀, and X + ½ when X < np₀.

📌 Example (Juvenile Court Orphans): A sociologist believes not more than 50% of children appearing in juvenile court 3+ times are orphans. Sample of 634 children: 341 are orphans. Test at 1% significance.

• Step 1: H₀: p ≤ 0.50, H₁: p > 0.50 (right-tailed)
• Step 2: α = 0.01
• Step 3: Z = [X - ½ - np₀] / √[np₀(1-p₀)] (using X - ½ since X=341 > np₀=317)
• Step 4: np₀ = 634(0.50) = 317. Z = (341 - 0.5 - 317) / √[634(0.50)(0.50)] = 23.5 / √158.5 = 23.5 / 12.59 = 1.87
• Step 5: Critical region: Z > 2.33
• Step 6: Since 1.87 < 2.33, do not reject H₀. Conclusion: The sociologist's hunch (not more than 50% are orphans) is acceptable.

⭐ Key Takeaways

The six-step hypothesis testing procedure is universal and must be applied systematically for every test. When comparing two population means with large samples (n₁, n₂ ≥ 30), use the Z-statistic, and unknown population variances can be replaced by sample variances S². For the two-tailed test, critical values are ±Z(α/2); for one-tailed tests, use Z(α) in the appropriate direction. When testing a population proportion p with large samples, apply the continuity correction of ±½ to improve the normal approximation, and the test statistic involves comparing the adjusted observed count to the expected count under H₀. Always compare the computed Z to the appropriate critical value from the standard normal table to determine if the null hypothesis should be rejected.

🧠 Quick Revision Questions

1. What are the six steps in the general hypothesis testing procedure, and why is it important to assume H₀ is true at the start?
2. For large samples, why can we replace σ² with S² in the two-sample Z-test, and what is the relationship between S² and s²?
3. In the computer analysts vs. nurses example, how was the computed Z value of 3.43 obtained, and why was H₀ rejected at α = 0.02?
4. When testing a population proportion, when do you use X - ½ versus X + ½ as the continuity correction, and why is this correction necessary?
5. In the juvenile court orphans example, what was the calculated Z value, why was it less than the critical value of 2.33, and what was the final conclusion about the sociologist's hunch?

📘 Lecture 39 — Hypothesis Testing Regarding p1-p2 (based on Z-statistic), The Student’s t-distribution, Confidence Interval for μ based on the t-distribution

📖 Overview: This lecture extends hypothesis testing to the comparison of two population proportions using the Z-statistic. It then introduces the Student’s t-distribution, explaining its properties and its application for constructing confidence intervals for the population mean when the population variance is unknown and the sample size is small.

🗂️ Topics Covered

This lecture begins by extending hypothesis testing to two population proportions via the Z-statistic, illustrated with examples comparing preferences between younger and older women and between educated and uneducated voters. It then formally introduces the Student’s t-distribution, detailing its mathematical definition and key properties, such as its shape and dependence on degrees of freedom. Finally, the lecture applies the t-distribution to construct a confidence interval for the population mean, demonstrating the procedure with a worked example involving the masses of ball bearings.

📝 Lecture Summary

Hypothesis Testing Regarding p1-p2 (based on Z-statistic)

This section extends hypothesis testing to compare the difference between two population proportions, (p_1 - p_2). The test statistic is a Z-statistic, which is valid for large sample sizes. When the null hypothesis states that the two proportions are equal (i.e., (p_1 - p_2 = 0)), the standard error of the difference in sample proportions is calculated using a pooled proportion. This pooled estimate, (\hat{p}_c), combines the successes from both samples and provides a better estimate of the common population proportion under the null hypothesis.

🔑 Definition — Pooled Proportion ((\hat{p}_c)): The weighted average of two sample proportions, used when the hypothesized difference between the two population proportions is zero. It is calculated as the total number of successes in both samples combined divided by the total number of observations in both samples. (\hat{p}_c = \frac{X_1 + X_2}{n_1 + n_2})

📐 Formula: Test Statistic for (p_1 - p_2) (when (H_0: p_1 - p_2 = 0)): [ Z = \frac{(\hat{p}_1 - \hat{p}_2) - 0}{\sqrt{\hat{p}_c \hat{q}_c \left(\frac{1}{n_1} + \frac{1}{n_2}\right)}} ] → The Z-score measures how many standard errors the observed sample difference is from the hypothesized difference of zero.

📌 Example: A perfume company tests if the proportion of younger women ((p_1)) who prefer its new perfume equals the proportion of older women ((p_2)) who prefer it. In a sample of 100 young women, 20 preferred it ((\hat{p}_1 = 0.20)). In a sample of 200 older women, 100 preferred it ((\hat{p}_2 = 0.50)). At (\alpha = 0.05), test (H_0: p_1 = p_2) vs. (H_1: p_1 \neq p_2).

1. Step 1: (H_0: p_1 = p_2) ((p_1 - p_2 = 0)), (H_1: p_1 \neq p_2) (two-tailed test).
2. Step 2: Level of significance (\alpha = 0.05).
3. Step 3: Test statistic is Z. Compute the pooled proportion (\hat{p}_c = \frac{20 + 100}{100 + 200} = \frac{120}{300} = 0.40). Then, (\hat{q}_c = 1 - 0.40 = 0.60).
4. Step 4: (Z = \frac{(0.20 - 0.50) - 0}{\sqrt{0.40 \times 0.60 \times \left(\frac{1}{100} + \frac{1}{200}\right)}} = \frac{-0.30}{\sqrt{0.24 \times 0.015}} = \frac{-0.30}{\sqrt{0.0036}} = \frac{-0.30}{0.06} = -5.00).
5. Step 5: Critical region for a two-tailed test at (\alpha = 0.05) is (Z < -1.96) or (Z > 1.96).
6. Step 6: The computed Z of -5.00 is less than -1.96, so it falls in the rejection region. Conclusion: Reject (H_0). There is sufficient evidence to conclude that the proportions of younger and older women who prefer the perfume are not equal.

💡 Why this matters: When the null hypothesis specifies a non-zero difference (e.g., (H_0: p_1 - p_2 > 0.10)), the standard error is calculated using the individual sample proportions ((\hat{p}_1) and (\hat{p}_2)) rather than a pooled proportion, as they are not estimating the same value under (H_0).

📌 Example: A candidate believes he appeals to at least 10% more of educated voters ((p_1)) than uneducated voters ((p_2)). In a sample of 100 educated voters, 62 support him ((\hat{p}_1 = 0.62)). In a sample of 150 uneducated voters, 69 support him ((\hat{p}_2 = 0.46)). At (\alpha = 0.05), test (H_0: p_1 - p_2 \geq 0.10) vs. (H_1: p_1 - p_2 < 0.10).

1. Step 1: (H_0: p_1 - p_2 \geq 0.10), (H_1: p_1 - p_2 < 0.10) (one-tailed test).
2. Step 2: (\alpha = 0.05).
3. Step 3: Test statistic is Z. The standard error is (\sqrt{\frac{\hat{p}_1 \hat{q}_1}{n_1} + \frac{\hat{p}_2 \hat{q}_2}{n_2}}).
4. Step 4: (Z = \frac{(0.62 - 0.46) - 0.10}{\sqrt{\frac{0.62 \times 0.38}{100} + \frac{0.46 \times 0.54}{150}}} = \frac{0.06}{\sqrt{0.002356 + 0.001656}} = \frac{0.06}{\sqrt{0.004012}} = \frac{0.06}{0.063} = 0.95).
5. Step 5: The critical region for a one-tailed test at (\alpha = 0.05) is (Z < -1.645).
6. Step 6: The calculated Z of 0.95 does not fall in the critical region. Conclusion: Do not reject (H_0). The data supports the candidate's claim.

The Student’s t-distribution

This section formally introduces the Student’s t-distribution, a probability distribution used when conducting statistical inference about a population mean when the population is normally distributed, the population variance (\sigma^2) is unknown, and the sample size is small (typically (n < 30)).

🔑 Definition — t-distribution: A bell-shaped, symmetric distribution that is flatter and has heavier tails than the standard normal distribution. Its exact shape is determined by a parameter called the degrees of freedom ((\nu)). The formula for the test statistic is: [ t = \frac{\bar{X} - \mu_0}{s / \sqrt{n}} ] where (s) is the sample standard deviation. This statistic follows a t-distribution with (\nu = n - 1) degrees of freedom.

Properties of the t-distribution:

1. It is bell-shaped and symmetric about (t = 0), ranging from (-\infty) to (\infty).
2. The shape is determined by the degrees of freedom; there is a distinct t-distribution for each value of (\nu).
3. For small (\nu), the t-distribution is flatter than the standard normal distribution (more variability).
4. As the degrees of freedom increase ((\nu \rightarrow \infty)), the t-distribution approaches the standard normal distribution.
5. The mean is zero (for (\nu \ge 2)).
6. The variance is (\sigma^2 = \nu / (\nu - 2)) for (\nu > 2), which is always greater than 1.
7. The distribution is unimodal with the mode at (t = 0).

Confidence Interval for μ based on the t-distribution

This section demonstrates how to construct a confidence interval for the population mean (\mu) when using the t-distribution.

📐 Formula: Confidence Interval for (\mu) using the t-distribution: [ \bar{X} \pm t_{\alpha/2, (n-1)} \left( \frac{s}{\sqrt{n}} \right) ] where (t_{\alpha/2, (n-1)}) is the critical value from the t-distribution with (n-1) degrees of freedom.

📌 Example: The masses (in grams) of 13 ball bearings are: 21.4, 23.1, 25.9, 24.7, 23.4, 24.5, 25.0, 22.5, 26.9, 26.4, 25.8, 23.2, 21.9. Calculate a 95% confidence interval for the mean mass ((\mu)) of the population, assumed normal.

1. Calculate sample statistics: (n = 13). Sample mean, (\bar{X} = 314.7 / 13 = 24.21) grams. Sample standard deviation, (s = \sqrt{37.43 / 12} = \sqrt{3.12} = 1.77) grams.
2. Find critical value: For a 95% confidence level, (\alpha = 0.05), so (\alpha/2 = 0.025). Degrees of freedom are (n-1 = 12). From the t-table, (t_{0.025, 12} = 2.179).
3. Calculate margin of error: (E = t_{\alpha/2, (n-1)} \times \frac{s}{\sqrt{n}} = 2.179 \times \frac{1.77}{\sqrt{13}} = 2.179 \times 0.491 = 1.07).
4. Construct confidence interval: (\bar{X} \pm E = 24.21 \pm 1.07 = (23.14, 25.28)).
5. Conclusion: We are 95% confident that the true population mean mass of the ball bearings lies between 23.14 grams and 25.28 grams.

⭐ Key Takeaways

• When testing the hypothesis (H_0: p_1 = p_2), a pooled proportion (\hat{p}_c) is used to estimate the common population proportion, and the Z-test formula incorporates this pooled estimate. If (H_0) specifies a non-zero difference, the standard error uses (\hat{p}_1) and (\hat{p}_2) individually.
• The t-distribution is the appropriate sampling distribution for inference about a population mean when the population is normally distributed but its variance is unknown and the sample size is small ((n < 30)).
• The t-distribution is symmetric and bell-shaped but has heavier tails and is more spread out than the standard normal distribution; its exact shape is determined by its degrees of freedom ((\nu = n-1)). As the degrees of freedom increase, it approaches the standard normal distribution.
• The formula for a confidence interval for (\mu) using the t-distribution is (\bar{X} \pm t_{\alpha/2, (n-1)} (s/\sqrt{n})), where the critical value (t_{\alpha/2, (n-1)}) is obtained from the t-table.
• The key difference between using the Z and t distributions lies in whether the population standard deviation (\sigma) is known (use Z) or unknown and estimated by the sample standard deviation (s) (use t, especially for small samples).

🧠 Quick Revision Questions

1. Under what conditions is the pooled proportion (\hat{p}_c) used in a two-proportion Z-test, and why?
2. State the formula for the test statistic used in hypothesis testing for (p_1 - p_2) when the null hypothesis specifies a non-zero difference.
3. List at least four properties of the Student’s t-distribution.
4. What are the necessary conditions for using the t-distribution to construct a confidence interval for the population mean (\mu)?
5. A sample of 10 items yields (\bar{X} = 50) and (s = 4). Find the critical t-value and calculate the 95% confidence interval for (\mu). (Hint: Use (n-1) = 9 degrees of freedom, and (t_{0.025, 9} = 2.262)).

📘 Lecture 40 — Tests and Confidence Intervals based on the t-distribution

📖 Overview: This lecture completes the discussion of statistical inference using the t-distribution, covering hypothesis testing for a single mean when the population variance is unknown and the sample is small. It then extends these concepts to compare two population means using independent small samples and paired observations, providing complete procedures for both confidence intervals and hypothesis tests.

🗂️ Topics Covered

The lecture covers hypothesis testing for a single population mean using the t-distribution with a detailed example, construction of confidence intervals for the difference between two means when variances are unknown but equal, hypothesis testing for the difference between two means with a worked example comparing rubber percentages, and finally the t-distribution approach for paired observations with an example on weight changes before and after training.

📝 Lecture Summary

Tests and Confidence Intervals based on the t-distribution

The lecture begins by reviewing the confidence interval for μ when drawing a small sample (n < 30) from a normal population with unknown variance σ². The confidence interval formula is:

x̄ ± t_(α/2, n-1) × (s/√n)

Where x̄ is the sample mean, s is the sample standard deviation, n is sample size, and t_(α/2, ν) is found from the t-table with ν = n-1 degrees of freedom. The lecture specifies the critical values: for 99% confidence, α/2 = 0.005; for 95% confidence, α/2 = 0.025; and for 90% confidence, α/2 = 0.05.

The discussion then moves to hypothesis testing regarding the mean of a normally distributed population with unknown σ² and small sample size (n < 30). This procedure is demonstrated through Example-1.

Example-1: A zoologist believes the average height of a particular animal species is 66 cm. A random sample of 10 animals yields heights: 63, 63, 66, 67, 68, 69, 70, 70, 71, 71 inches. Test the hypothesis at α = 0.05.

Solution: i) H₀: μ = 66 vs H₁: μ ≠ 66 (two-tailed test) ii) Significance level α = 0.05 iii) Test statistic: t = (x̄ - μ₀)/(s/√n), which under H₀ follows t-distribution with n-1 = 9 degrees of freedom iv) Calculations: Σx_i = 678, so x̄ = 67.8 inches; Σx_i² = 46050; s² = (1/9)[46050 - (678)²/10] = 9.0667; s = 3.01 inches t = (67.8 - 66)/(3.01/√10) = (1.8 × 3.1623)/3.01 = 1.89 v) Critical region for two-tailed test: |t| > t_(0.025,9) = 2.262 vi) Conclusion: Since computed t = 1.89 does not fall in the critical region, we do not reject H₀ and conclude the mean height is 66 centimeters.

💡 Why this matters: This illustrates the complete hypothesis testing procedure using t-distribution, emphasizing that the test statistic must exceed the critical value from the t-table to reject the null hypothesis.

Confidence Interval for μ₁ - μ₂ with Small Samples and Equal Variances

Next, the lecture considers constructing a confidence interval for μ₁ - μ₂ when drawing small samples from two normally distributed populations having unknown but equal variances. This uses the pooled variance s_p².

Example: A record company executive wants to estimate the difference in average play-length between pop music and semi-classical songs. She randomly selects 10 semi-classical songs and 9 pop songs, with play-lengths given in minutes. Calculate a 99% confidence interval.

Solution: Degrees of freedom: n₁ + n₂ - 2 = 10 + 9 - 2 = 17 t-table value for 99% confidence: t_(0.005, 17) = 2.898

Calculations: Semi-Classical: n₁ = 10, x̄₁ = 3.465, s₁ = 0.3575 Pop Music: n₂ = 9, x̄₂ = 4.064, s₂ = 0.2417

Pooled variance: s_p² = [(n₁-1)s₁² + (n₂-1)s₂²]/(n₁+n₂-2) = [9(0.3575)² + 8(0.2417)²]/17 = [1.1503 + 0.4674]/17 = 1.6177/17 = 0.0952 s_p = √0.0952 = 0.31

Confidence interval: (x̄₁ - x̄₂) ± t_(α/2, ν) × s_p × √(1/n₁ + 1/n₂) = (3.465 - 4.064) ± (2.898)(0.31)√(1/10 + 1/9) = -0.599 ± (2.898)(0.31)(0.36) = -0.599 ± 0.411 Therefore: -1.010 ≤ μ₁ - μ₂ ≤ -0.188

With 99% confidence, the true difference is between -1.01 and -0.188 minutes. Since zero is not in this interval, we conclude there is a significant difference in average play-length, with pop songs being longer.

Hypothesis Testing for μ₁ - μ₂ with Small Samples and Equal Variances

Example: From an area planted with guayule, 54 plants were selected. Of these, 15 were "off types" and 12 were "aberrant." Rubber percentages are given. Test the hypothesis that the mean rubber percentage of Aberrants is at least 1 percent more than off types. Use α = 0.05, assuming normal populations with equal variances.

Let subscript 1 = Aberrants, subscript 2 = off types.

Solution: i) H₀: μ₁ - μ₂ ≥ 1 vs H₁: μ₁ - μ₂ < 1 (left-tailed test) ii) α = 0.05 iii) Test statistic: t = [(x̄₁ - x̄₂) - (μ₁ - μ₂)₀] / [s_p × √(1/n₁ + 1/n₂)], with ν = n₁ + n₂ - 2 = 25 degrees of freedom iv) Computations: x̄₁ = Σx₁/n₁ = 80.92/12 = 6.74 x̄₂ = Σx₂/n₂ = 84.25/15 = 5.62 Σ(x₁ - x̄₁)² = Σx₁² - (Σx₁)²/n₁ = 561.6402 - (80.92)²/12 = 561.6402 - 545.6705 = 15.9697 Σ(x₂ - x̄₂)² = Σx₂² - (Σx₂)²/n₂ = 478.9779 - (84.25)²/15 = 478.9779 - 473.2042 = 5.7737 s_p² = [15.9697 + 5.7737]/(12 + 15 - 2) = 21.7434/25 = 0.8697 s_p = √0.8697 = 0.93 t = [(6.74 - 5.62) - 1] / [0.93 × √(1/12 + 1/15)] = 0.12 / [0.93 × 0.36] = 0.12/0.333 = 0.33 v) Critical region for left-tailed test: t < -t_(0.05, 25) i.e. t < -1.708 vi) Conclusion: Since computed t = 0.33 falls in the acceptance region, we accept H₀. The mean rubber percentage of Aberrants is at least 1 percent more than off types.

🔑 Definition — Pooled Variance (s_p²): When two populations have equal but unknown variances, the best estimate combines information from both samples: s_p² = [(n₁-1)s₁² + (n₂-1)s₂²]/(n₁+n₂-2)

T-Distribution in the Case of Paired Observations

The lecture introduces paired observations, where the two samples are not independent because observations come in pairs that are naturally related. Natural pairing occurs when measurements are taken on the same unit at two different times, such as "before" and "after" measurements on the same individual.

For paired data, we compute the difference (d) between the two observations of each pair. The test statistic for testing whether the mean difference (μ_d) equals zero is:

t = (d̄ - μ_d) / (s_d/√n) = d̄ / (s_d/√n) when testing μ_d = 0

Where d̄ is the mean of the differences, s_d is the standard deviation of the differences, and n is the number of pairs.

Example: Ten young recruits underwent strenuous physical training. Their weights were recorded before and after training: Recruits 1-10: Before (125,195,160,171,140,201,170,176,195,139) and After (136,201,158,184,145,195,175,190,190,145) Using α = 0.05, determine if the program affects average weight. Assume normal distribution.

The test statistic is t = d̄/(s_d/√n), where d = weight after - weight before for each recruit.

⭐ Key Takeaways

The most critical concepts from this lecture are: the t-distribution is used for inference about means when σ² is unknown and n < 30, requiring normal population assumption. For comparing two independent means with unknown but equal variances, the pooled variance estimate combines both sample variances. The confidence interval for μ₁-μ₂ uses the t-distribution with n₁+n₂-2 degrees of freedom. Hypothesis testing follows the same logic but compares computed t to critical values. For paired observations, compute differences within each pair and test using the sample of differences with n-1 degrees of freedom. The key difference between independent and paired samples is that pairing reduces variability by controlling for individual differences, making it a more powerful test when applicable.

🧠 Quick Revision Questions

1. What are the three conditions that require using the t-distribution instead of the normal distribution for inference about a population mean?
2. In Example-1 (heights of animals), why was the conclusion "do not reject H₀" reached when the sample mean (67.8) was different from the hypothesized mean (66)?
3. What is the formula for the pooled variance s_p², and when is it appropriate to use it?
4. In the record company example, why did the confidence interval (-1.010, -0.188) lead to the conclusion that there is a significant difference between pop and semi-classical songs?
5. What is the fundamental difference between independent samples and paired samples, and how does this affect the degrees of freedom for the t-test?

📘 Lecture 41 — Hypothesis-Testing Regarding Two Population Means in the Case of Paired Observations (t-distribution) • The Chi-square Distribution • Hypothesis Testing and Interval Estimation Regarding a Population Variance (based on Chi-square Distribution)

📖 Overview: This lecture continues the discussion of hypothesis-testing for two population means using paired observations, where natural or designed pairing eliminates unwanted variation. It then introduces the chi-square distribution, its properties, and its application in interval estimation and hypothesis-testing regarding a single population variance.

🗂️ Topics Covered

The lecture covers the paired difference t-test procedure with detailed examples, including computation of differences, test statistics, and confidence intervals. It then formally defines the chi-square distribution, lists its key properties (shape, mean, variance, moments), and demonstrates how to use chi-square tables for constructing confidence intervals and conducting hypothesis tests for population variance under normality assumptions.

📝 Lecture Summary

Hypothesis-Testing regarding Two Population Means in the Case of Paired Observations (t-distribution)

In many situations, pairing occurs naturally (e.g., before/after measurements on the same subject) or by design (e.g., pairing plots by soil type to eliminate extraneous effects). By taking observations in pairs, we remove undesirable sources of variation. The procedure treats the differences between paired observations as a random sample from a normal population with mean μd = μ1 – μ2 and unknown standard deviation σd, then performs a one-sample t-test on these differences. This is called a paired difference t-test or paired t-test.

🔑 Definition — Paired t-test: A hypothesis test that compares two population means by analyzing the differences between paired observations, treating these differences as a single sample.

📐 Formula: t = (d̄ – 0) / (sd / √n) = d̄ / (sd / √n) → The test statistic follows a t-distribution with n – 1 degrees of freedom.

📌 Example: Ten recruits had weights recorded before and after training. Data: Recruit 1: before 125, after 136 (d=11); Recruit 2: 195, 201 (d=6); Recruit 3: 160, 158 (d=-2); Recruit 4: 171, 184 (d=13); Recruit 5: 140, 145 (d=5); Recruit 6: 201, 195 (d=-6); Recruit 7: 170, 175 (d=5); Recruit 8: 176, 190 (d=14); Recruit 9: 195, 190 (d=-5); Recruit 10: 139, 145 (d=6). Σd = 47, Σd² = 673, n = 10. d̄ = 47/10 = 4.7. sd² = [673 – (47)²/10] / 9 = [673 – 220.9] / 9 = 452.1/9 = 50.23, so sd = √50.23 = 7.09. t = 4.7 / (7.09/√10) = (4.7 × 3.16) / 7.09 = 14.852 / 7.09 = 2.09. Critical region: |t| ≥ t0.025(9) = 2.262. Since 2.09 < 2.262, accept H0: μd = 0. Conclusion: Data do not provide sufficient evidence that the programme affects average weight.

Hypothesis-Testing Procedure Regarding Two Populations Means in the Case of Paired Observations

When observations from two samples are paired (naturally or by design), compute the difference d = x1 – x2 for each pair. The sample mean of differences is d̄ = Σd / n, and the sample standard deviation is sd = √[Σ(d – d̄)² / (n – 1)]. Under the assumptions that (1) differences form a random sample and (2) differences are normally distributed, the test statistic t = (d̄ – 0) / (sd/√n) follows a t-distribution with ν = n – 1 degrees of freedom. Testing H0: μ1 = μ2 is equivalent to testing H0: μd = 0.

📌 Example: Two varieties of wheat planted in 10 localities. Variety I: 45, 32, 58, 57, 60, 38, 47, 51, 42, 38. Variety II: 47, 34, 60, 59, 63, 44, 49, 53, 46, 41. Differences (I – II): -2, -2, -2, -2, -3, -6, -2, -2, -4, -3. Σd = -28, Σd² = 94, n = 10. d̄ = -28/10 = -2.8. sd² = [94 – (-28)²/10] / 9 = [94 – 784/10] / 9 = [94 – 78.4] / 9 = 15.6/9 = 1.7333, sd = √1.7333 = 1.32. t = -2.8 / (1.32/√10) = (-2.8 × 3.1623) / 1.32 = -8.85 / 1.32 = -6.71. Critical region (one-tailed, α=0.05): t < t0.05(9) = -1.833. Since -6.71 < -1.833, reject H0. Conclusion: Mean yield of variety-1 is less than variety-2. 90% confidence interval for μd: d̄ ± t0.05,9 × sd/√n = -2.8 ± 1.833 × (1.32/3.1623) = -2.8 ± 1.833 × 0.417 = -2.8 ± 0.765 → (-3.565, -2.035). 💡 Why this matters: Pairing by design (e.g., same locality for both varieties) eliminates extraneous factors like soil fertility, giving a more accurate comparison of the fertilizers themselves.

The Chi-square (χ²) Distribution

The chi-square distribution is a continuous distribution ranging from 0 to +∞ with one parameter ν (degrees of freedom). Its probability density function is f(x) = [1 / (2^(ν/2) Γ(ν/2))] × x^((ν/2)–1) × e^(–x/2), for 0 < x < ∞.

🔑 Definition — Chi-square distribution: A family of continuous probability distributions used primarily in statistical inference about population variances.

Properties of the Chi-Square Distribution:

1. It is a continuous distribution ranging from 0 to +∞.
2. The shape is determined by ν (degrees of freedom) — positively skewed, with skewness decreasing as ν increases.
3. The curve tends to the normal distribution as ν approaches infinity.
4. Mean = ν, Variance = 2ν.
5. Moments about the origin: μ'₁ = ν, μ'₂ = ν(ν+2), μ'₃ = ν(ν+2)(ν+4), μ'₄ = ν(ν+2)(ν+4)(ν+6). Moment ratios: β₁ = 8/ν, β₂ = 3 + 12/ν.

Interval Estimation Regarding Population Variance (Based on Chi-square Distribution)

If X̄ and S² are the mean and variance of a random sample of size n from a normal population with variance σ², then the statistic χ² = Σ(Xi – X̄)² / σ² = nS² / σ² = (n-1)s² / σ² follows a chi-square distribution with (n – 1) degrees of freedom. The 100(1–α)% confidence interval for σ² is:

📐 Formula: Σ(Xi – X̄)² / χ²α/2, (n-1) < σ² < Σ(Xi – X̄)² / χ²1-α/2, (n-1)

📌 Example: Sample of n=8 students' marks: 9, 14, 10, 12, 7, 13, 11, 12. X̄ = (9+14+10+12+7+13+11+12)/8 = 88/8 = 11. Σ(Xi – X̄)² = (9-11)² + (14-11)² + (10-11)² + (12-11)² + (7-11)² + (13-11)² + (11-11)² + (12-11)² = 4+9+1+1+16+4+0+1 = 36. For 90% confidence (α=0.10), from chi-square table with 7 df: χ²0.05(7) = 14.07, χ²0.95(7) = 2.17. 90% CI for σ²: 36/14.07 < σ² < 36/2.17 → 2.56 < σ² < 16.61. Taking square roots: 1.6 < σ < 4.1. 💡 Why this matters: The confidence interval for variance is very wide with small samples, showing that σ cannot be estimated with much precision for n < 30.

Hypothesis Testing Regarding Population Variance (Based on Chi-square Distribution)

To test H0: σ² = σ²0 against H1: σ² > σ²0 (or other alternatives), compute χ² = nS² / σ²0. Under H0, this follows a chi-square distribution with (n–1) degrees of freedom, assuming normality.

📌 Example: Test H0: σ² = 25 vs H1: σ² > 25, with n=16, S²=31.5, α=0.05. χ² = (16 × 31.5) / 25 = 504/25 = 20.16. Critical region: χ² > χ²0.05(15) = 25.00. Since 20.16 < 25.00, accept H0. Conclusion: Reasonable evidence that σ² = 25.

⭐ Key Takeaways

The paired t-test treats differences as a single sample, testing H0: μd = 0 against various alternatives. The chi-square distribution is positively skewed, with mean = ν and variance = 2ν; its shape depends only on degrees of freedom. For inference on population variance (σ²) from a normal population, use χ² = (n-1)s²/σ² with (n-1) df. Confidence intervals for σ² are computed using chi-square critical values, and hypothesis tests compare the observed chi-square value to the upper or lower percentage points from the chi-square table. With small samples, variance estimation has wide confidence intervals and limited precision.

🧠 Quick Revision Questions

1. What are the two assumptions required for the paired difference t-test?
2. How do you compute the test statistic for testing H0: μd = 0?
3. What is the mean and variance of a chi-square distribution with ν degrees of freedom?
4. Write the formula for a 90% confidence interval for population variance σ².
5. In hypothesis testing for variance, what critical region would you use for a one-tailed test where H1: σ² > σ²0 at α = 0.05?

📘 Lecture 42 — The F-Distribution

📖 Overview: This lecture introduces the F-distribution, its mathematical definition and properties, and demonstrates its use in statistical inference for comparing variances of two normal populations. It covers hypothesis testing and confidence interval estimation for the ratio of two population variances, building on earlier chi-square based tests for a single variance.

🗂️ Topics Covered

The lecture covers the F-distribution's mathematical definition and properties, followed by its application in constructing confidence intervals for the variance ratio of two normal populations. Hypothesis testing for equality of variances using the F-test is demonstrated with worked examples, including a test for greater variability between locations versus times in plankton data and a test for a specific variance ratio (1.25) for method comparision.

📝 Lecture Summary

The F-Distribution

The mathematical equation of the F-distribution is: $$f(x) = \frac{\Gamma[(v_1 + v_2)/2] (v_1/v_2)^{v_1/2} x^{(v_1/2)-1}}{\Gamma(v_1/2) \Gamma(v_2/2)[1 + (v_1 x / v_2)]^{(v_1+v_2)/2}}, \quad 0 < x < \infty$$

This distribution has two parameters, $v_1$ and $v_2$, which are the degrees of freedom of the F-distribution. It is usually abbreviated as $F(v_1, v_2)$, where $v_1$ is numerator degrees of freedom and $v_2$ is denominator degrees of freedom.

PROPERTIES OF F-DISTRIBUTION

1. The F-distribution is a continuous distribution ranging from zero to plus infinity.
2. The curve of the F-distribution is positively skewed. As the degrees of freedom $v_1$ and $v_2$ become large, the F-distribution approaches the normal distribution.
3. For $v_2 > 2$, the mean of the F-distribution is $\frac{v_2}{v_2 - 2}$, which is greater than 1.
4. For $v_2 > 4$, the variance of the F-distribution is $\sigma^2 = \frac{2v_2^2(v_1 + v_2 - 2)}{v_1(v_2 - 2)^2 (v_2 - 4)}$.
5. The F-distribution for $v_1 > 2, v_2 > 2$ is unimodal, with the mode at $\frac{v_2(v_1 - 2)}{v_1(v_2 + 2)}$, which is always less than 1.
6. If $F$ has an F-distribution with $v_1$ and $v_2$ degrees of freedom, then the reciprocal has an F-distribution with $v_2$ and $v_1$ degrees of freedom.

Tables of the F-distribution are provided for 5%, 2½%, and 1% right-tail areas. Separate tables exist for each significance level, with columns representing numerator degrees of freedom ($v_1$) and rows representing denominator degrees of freedom ($v_2$).

💡 Why this matters: The reciprocal property is crucial for obtaining lower-tail critical values from tables that only give upper-tail values—simply take the reciprocal of the corresponding value with degrees of freedom swapped.

Hypothesis Testing and Interval Estimation for Comparing Two Population Variances

CONFIDENCE INTERVAL FOR THE VARIANCE RATIO $\sigma_1^2 / \sigma_2^2$

Let two independent random samples of size $n_1$ and $n_2$ be taken from two normal populations with variances $\sigma_1^2$ and $\sigma_2^2$, and let $s_1^2$ and $s_2^2$ be the unbiased estimators of $\sigma_1^2$ and $\sigma_2^2$.

Then, the quantity $\frac{s_1^2 / \sigma_1^2}{s_2^2 / \sigma_2^2}$ has an F-distribution with $(n_1 - 1, n_2 - 1)$ degrees of freedom.

The confidence interval for $\sigma_1^2 / \sigma_2^2$ is given by: $$\left[ \frac{s_1^2}{s_2^2} \cdot \frac{1}{F_{\alpha/2}(n_1 - 1, n_2 - 1)}, \quad \frac{s_1^2}{s_2^2} \cdot F_{\alpha/2}(n_2 - 1, n_1 - 1) \right]$$

We can also find a confidence interval for $\sigma_1 / \sigma_2$ by taking the square root of the endpoints.

🔑 Definition — F-distribution for variance ratio: If $s_1^2$ and $s_2^2$ are unbiased sample variances from two independent normal populations, then $\frac{s_1^2 / \sigma_1^2}{s_2^2 / \sigma_2^2} \sim F(n_1 - 1, n_2 - 1)$.

📌 Example: A random sample of 12 salt-water fish had standard deviation $s_1 = 2.3$ inches. A sample of 10 fresh-water fish had $s_2 = 1.5$ inches. Find a 90% confidence interval for $\sigma_1^2 / \sigma_2^2$.

Solution: $s_1^2 = (2.3)^2 = 5.29$, $s_2^2 = (1.5)^2 = 2.25$ $n_1 - 1 = 11$, $n_2 - 1 = 9$ For 90% CI, $\alpha = 0.10$, so $\alpha/2 = 0.05$ $F_{0.05}(11, 9) = 3.1$ (obtained by interpolation: 3.23 for (8,9) and 3.07 for (12,9), mean of these with (10,9) = 3.15, mean with (11,9) = 3.11 ≈ 3.1) $F_{0.05}(9, 11) = 2.9$

90% CI = $\left[ \frac{5.29}{2.25} \cdot \frac{1}{3.1}, \frac{5.29}{2.25} \cdot 2.9 \right] = [0.76, 6.81]$

Taking square root, 90% CI for $\sigma_1 / \sigma_2$ is $(0.87, 2.61)$.

HYPOTHESIS TESTING FOR EQUALITY OF TWO VARIANCES

The test statistic for $H_0: \sigma_1^2 = \sigma_2^2$ is $F = \frac{s_1^2}{s_2^2}$ (with $s_1^2$ being the larger variance, typically placed in the numerator for a one-tailed test).

📌 Example: Series I (same place, 10 hauls): 80, 96, 102, 77, 97, 110, 99, 88, 103, 108. Series II (different places, 7 hauls): 74, 122, 92, 81, 104, 92, 92. Test at $\alpha = 0.05$ whether variability between places exceeds variability between times.

Solution: Step 1: $H_0: \sigma_1^2 \ge \sigma_2^2$ (time variability ≥ place variability); $H_A: \sigma_1^2 < \sigma_2^2$ (greater variability between places) Step 2: $\alpha = 0.05$ Step 3: Test statistic $F = \frac{s_2^2}{s_1^2}$ with $(n_2 - 1, n_1 - 1)$ degrees of freedom Step 4: Computations: $\sum X_1 = 960$, $\sum X_1^2 = 93276$, $n_1 = 10$ $s_1^2 = \frac{1}{9}[93276 - (960)^2/10] = \frac{1}{9}[93276 - 92160] = \frac{1116}{9} = 124$ $\sum X_2 = 657$, $\sum X_2^2 = 63129$, $n_2 = 7$ $s_2^2 = \frac{1}{6}[63129 - (657)^2/7] = \frac{1}{6}[63129 - 61664.14] = \frac{1464.86}{6} = 244.14$ $F = \frac{244.14}{124} = 1.97$ Step 5: Critical region (right-tailed): $F > F_{0.05}(6, 9) = 3.37$ Step 6: Since $1.97 < 3.37$, do not reject $H_0$. There is insufficient evidence of greater variability between different places.

TESTING FOR A SPECIFIC VARIANCE RATIO

If $H_0: \sigma_1^2 / \sigma_2^2 = k$, the test statistic is: $$F = \frac{s_1^2}{k \cdot s_2^2}$$ which has an F-distribution with $(n_1 - 1, n_2 - 1)$ degrees of freedom.

📌 Example: Method I variability should not exceed 125% of Method II variability. Sample: $n_1 = n_2 = 21$, $\sum (X_1 - \bar{X}_1)^2 = 720$, $\sum (X_2 - \bar{X}_2)^2 = 340$.

$H_0: \sigma_1^2 \le 1.25\sigma_2^2$ vs $H_1: \sigma_1^2 > 1.25\sigma_2^2$ $s_1^2 = 720/20 = 36$, $s_2^2 = 340/20 = 17$ $F = \frac{36}{1.25 \times 17} = \frac{36}{21.25} = 1.694$ Critical value: $F_{0.05}(20, 20) = 2.12$ (from table) Since $1.694 < 2.12$, do not reject $H_0$; Method I can be recommended.

⭐ Key Takeaways

The F-distribution is a continuous, positively skewed distribution used to compare two population variances by forming the ratio of their sample variances. Confidence intervals for the variance ratio require F-values from tables at $\alpha/2$, using interpolation if needed for missing degrees of freedom. In hypothesis testing, the ratio of the larger sample variance to the smaller one is the test statistic, and the alternative hypothesis determines whether to use a one-tailed or two-tailed test. When testing for a specific ratio $k$, the test statistic becomes $s_1^2/(k \cdot s_2^2)$, still following an F-distribution. Always verify the normality assumption before applying these procedures.

🧠 Quick Revision Questions

1. What are the two parameters of the F-distribution called, and how do they relate to sample sizes?
2. What property relates $F_{1-\alpha}(v_1, v_2)$ to $F_\alpha(v_2, v_1)$?
3. In constructing a 90% confidence interval for $\sigma_1^2/\sigma_2^2$, which F-values are needed, and with what degrees of freedom?
4. In the plankton example, why was the test statistic calculated as $s_2^2/s_1^2$ rather than $s_1^2/s_2^2$?
5. How does the test statistic change when testing $H_0: \sigma_1^2/\sigma_2^2 = 1.25$ compared to testing $H_0: \sigma_1^2 = \sigma_2^2$?

📘 Lecture 43 — Analysis of Variance & Experimental Design

📖 Overview: This lecture introduces Analysis of Variance (ANOVA), a statistical procedure for testing the equality of several population means simultaneously, which avoids the inefficiency of conducting multiple two-sample t-tests. It also covers the fundamental concepts of Experimental Design, specifically the Completely Randomized (CR) Design and the Randomized Complete Block (RCB) Design, providing a step-by-step example of performing ANOVA for a CR design.

🗂️ Topics Covered

The lecture begins by explaining why multiple two-sample t-tests are inappropriate for comparing more than two means and introduces ANOVA as a solution. It defines experimental design and its components, then distinguishes between systematic and randomized designs. The Completely Randomized (CR) Design is explained in detail, including its definition, assumptions, and the construction and interpretation of the ANOVA table. The lecture concludes by introducing the Randomized Complete Block (RCB) Design and presenting an example for analysis.

📝 Lecture Summary

Analysis of Variance

Earlier, we compared two-population means using a two-sample t-test. However, comparing more than two population means simultaneously requires a simpler procedure. For k populations, there are k(k-1)/2 possible pairwise comparisons. For example, comparing 4 means requires 6 separate tests, and comparing 10 means requires 45 tests. This is tedious and time-consuming. The Analysis of Variance (ANOVA) , introduced by Sir R.A. Fisher, is a procedure that enables us to test the hypothesis of equality of several population means (H0: μ1 = μ2 = μ3 = ... = μk against HA: not all the means are equal) simultaneously.

Experimental Design

An experimental design is a plan used to collect data relevant to a problem in a way that provides a basis for valid and objective inference. The plan typically includes the selection of treatments, the specification of the experimental layout, and the assignment of treatments to experimental units.

There are two types of designs: Systematic and Randomized. This course focuses on randomized designs, for which analysis is carried out through Analysis of Variance. Two basic randomized designs are the Completely Randomized (CR) Design and the Randomized Complete Block (RCB) Design.

The Completely Randomized Design (CR Design)

A Completely Randomized (CR) design is the simplest type of basic design where treatments are assigned to experimental units completely at random, without any restrictions. This design is applicable when the entire experimental material is homogeneous (i.e., all experimental units are similar).

Example: An experiment was conducted to compare the yields of three varieties of potatoes (A, B, C). Each variety was assigned at random to equal-size plots, four times. The yields were as follows:

Solution (Hypothesis-Testing Procedure): i) H0: μA = μB = μC ; HA: Not all three means are equal. ii) Level of Significance: α = 0.05 iii) Test Statistic: F = MSTreatments / MSError, which follows an F-distribution with ν1 = k-1 and ν2 = n-k.

Computation of the ANOVA Table: The ANOVA table for CR Design has columns for Source of Variation, d.f., Sum of Squares, Mean Square, and F.

• Sources of Variation:
  • Total: The overall variation in all data values.
  • Between treatments (SST): Variability between the means of the different treatments (e.g., different potato varieties).
  • Within treatments (Error) (SSE): Variability within the data of the same treatment. This is also called error variation.
• Formulae:
  • Correction Factor (C.F.): C.F. = T..² / n
  • Total Sum of Squares (TSS): TSS = ΣᵢΣⱼ Xᵢⱼ² - C.F.
  • Treatment Sum of Squares (SST): SST = Σⱼ (T.ⱼ² / r) - C.F.
  • Error Sum of Squares (SSE): SSE = TSS - SST
  • Degrees of Freedom: Total (n-1) = Treatment (k-1) + Error (n-k)
• Calculations:
  • ΣᵢΣⱼ Xᵢⱼ² = 4953
  • T.. = 237
  • Σ T.ⱼ² = 18941
  • C.F. = (237)² / 12 = 4680.75
  • TSS = 4953 - 4680.75 = 272.25
  • SST = (18941 / 4) - 4680.75 = 4735.25 - 4680.75 = 54.50
  • SSE = 272.25 - 54.50 = 217.75
  • MST = SST / (k-1) = 54.50 / 2 = 27.25
  • MSE = SSE / (n-k) = 217.75 / 9 = 24.19
  • Computed F = MST / MSE = 27.25 / 24.19 = 1.13

The completed ANOVA Table is:

v) Critical Region: For a right-tailed test: F > Fα (k-1, n-k). Here, F > F0.05 (2,9) = 4.26 vi) Conclusion: Since the computed F (1.13) does not fall in the critical region, we accept H0. There is no significant difference among the yielding capabilities of the three potato varieties.

💡 Why this matters: The ANOVA technique is valid under these assumptions:

• The k populations are normally distributed.
• All k populations have equal variances (homoscedasticity).
• The k samples are drawn randomly and independently.

The Randomized Complete Block Design (RCB Design)

A Randomized Complete Block (RCB) Design is used when the experimental material is not homogeneous overall. The material is divided into groups or blocks such that units within a block are relatively homogeneous. Each block contains a complete set of treatments, and treatments are allocated at random within each block. This design brings variability under control.

Example (for next lecture): In a feeding experiment, four types of rations (A, B, C, D) were given to animals in five groups (I, II, III, IV, V). The results (gain in weight in pounds) were:

⭐ Key Takeaways

• Analysis of Variance (ANOVA) is the appropriate procedure for testing the equality of three or more population means, avoiding the inefficiency and inflated error rate of multiple two-sample t-tests.
• The Completely Randomized (CR) Design requires homogeneous experimental units and uses One-Way ANOVA to partition the total variation into variation between treatments (SST) and variation within treatments (Error, SSE) .
• The F-statistic is computed as MST/MSE. A larger F value indicates greater evidence against the null hypothesis of equal means.
• The ANOVA table is a structured way to organize the sums of squares, degrees of freedom, and mean squares needed to calculate the F-statistic. The total sum of squares equals the sum of treatment and error sums of squares.
• The Randomized Complete Block (RCB) Design is used when experimental units are not homogeneous; it groups them into blocks to control for extraneous variability and uses Two-Way ANOVA for analysis.

🧠 Quick Revision Questions

1. What is the main disadvantage of using multiple two-sample t-tests to compare more than two population means, and how does Analysis of Variance (ANOVA) solve this problem?
2. What are the key assumptions underlying the validity of the ANOVA technique for the Completely Randomized Design?
3. In the ANOVA table for a CR design, what does the "Error" or "Within treatments" source of variation represent, and why is it important for the F-test?
4. How is the test statistic F computed in a one-way ANOVA, and what does a large computed F-value suggest about the population means?
5. In what practical scenario would a researcher choose a Randomized Complete Block (RCB) Design over a Completely Randomized (CR) Design?

📘 Lecture 44 — Randomized Complete Block Design, The Least Significant Difference (LSD) Test, Chi-Square Test of Goodness of Fit

📖 Overview: This lecture completes the analysis of the Randomized Complete Block (RCB) Design with a detailed example, teaching how to partition total variation into treatment, block, and error components. It then introduces the Least Significant Difference (LSD) test for pairwise comparisons of treatment means after a significant ANOVA result. Finally, it covers the Chi-Square Test of Goodness of Fit, demonstrating how to mathematically determine whether observed frequencies follow a hypothesized theoretical distribution.

🗂️ Topics Covered

The lecture covers the complete analysis of variance for a Randomized Complete Block Design using a feeding experiment example with four rations and five groups of animals. It then explains the Least Significant Difference (LSD) test procedure for comparing all pairs of treatment means simultaneously. The final major topic is the Chi-Square Test of Goodness of Fit, including its hypothesis testing procedure, computation of expected frequencies from a fitted binomial distribution, and the requirement for merging categories with small expected frequencies.

📝 Lecture Summary

Randomized Complete Block Design

The lecture begins with a detailed example of an RCB design involving a feeding experiment where four types of rations (treatments) were given to animals in five groups (blocks) of four each. The data represents weight gains in pounds, and the analysis of variance aims to test two sets of hypotheses. The primary hypothesis tests whether all ration means are equal, while a secondary hypothesis tests whether all group (block) means are equal.

🔑 Definition — Randomized Complete Block Design: An experimental design where subjects are first divided into homogeneous groups (blocks), and then all treatments are randomly assigned within each block.

Computations: The analysis requires calculating Total Sum of Squares, Treatment Sum of Squares, Block Sum of Squares, and Error Sum of Squares.

📐 Total SS = ΣΣX²ij − (T..²/n) = 21725.22 − (656.4²/20) = 21725.22 − 21543.05 = 182.17

📐 Treatment SS = Σ(T.j²/r) − (T..²/n) = 108387.48/5 − 21543.05 = 21677.50 − 21543.05 = 134.45

📐 Block SS = Σ(Bi.²/c) − (T..²/n) = 86258.04/4 − 21543.05 = 21564.51 − 21543.05 = 21.46

📐 Error SS = Total SS − (Treatment SS + Block SS) = 182.17 − (134.45 + 21.46) = 26.26

The degrees of freedom are: treatments (c-1=3), blocks (r-1=4), error [(r-1)(c-1)=12], and total (rc-1=19).

The ANOVA table shows F1 = 44.82/2.19 = 20.47 for treatments and F2 = 5.36/2.19 = 2.45 for blocks.

Conclusions: Since F1 = 20.47 exceeds the critical value F0.05(3,12) = 3.49, we reject H0 and conclude there is a difference among at least two treatment means. Since F2 = 2.45 does not exceed F0.05(4,12) = 3.26, we accept H′0 and conclude that blocking was not actually required for this experiment.

💡 Why this matters: If blocking is unnecessary, future experiments can use a Completely Randomized Design, which retains more degrees of freedom for Error, providing a better estimate of error variation.

The Least Significant Difference (LSD) Test

After finding significant differences among treatment means, the LSD test helps determine which specific treatments differ from each other. This procedure computes the smallest difference that would be judged significant and compares all pairwise differences of means with this value.

🔑 Definition — Least Significant Difference (LSD): The smallest absolute difference between two sample means that would be considered statistically significant.

📐 Formula: LSD = tα/2(ν) × √[2(MSE)/r] → Where MSE is the Mean Square for Error, r is the equal sample size, and tα/2(ν) is the t-value at α/2 level with error degrees of freedom.

📌 Example: With MSE = 2.19, r = 5, and t0.025(12) = 2.179: LSD = 2.179 × √[2(2.19)/5] = 2.179 × √0.876 = 2.179 × 0.936 = 2.04

The four treatment means are: X̄D = 28.38, X̄C = 33.70, X̄A = 34.42, X̄B = 34.78.

Arranging in ascending order and comparing differences against LSD = 2.04:

• |X̄C − X̄D| = |33.70 − 28.38| = 5.32 > 2.04 → Significant
• |X̄A − X̄C| = |34.42 − 33.70| = 0.72 < 2.04 → Not significant
• |X̄B − X̄A| = |34.78 − 34.42| = 0.36 < 2.04 → Not significant

The conclusion is that rations C, A, and B are not significantly different from each other, but ration D is significantly different and performs poorest.

Chi-Square Test of Goodness of Fit

This test determines mathematically whether observed frequencies from a sample match expected frequencies based on a theoretical distribution. The lecture demonstrates this using the binomial distribution example from Lecture No. 28, where a loaded die was tossed 5 times and the number of sixes was recorded.

Properties of Chi-Square Distribution:

1. Continuous distribution ranging from 0 to +∞
2. Positively skewed, with skewness decreasing as degrees of freedom increase
3. Tends to normal distribution as degrees of freedom approach infinity

Hypothesis Testing Procedure for Goodness of Fit:

Step 1: H0: The fit is good vs HA: The fit is not good Step 2: α = 0.05 Step 3: Test-statistic: χ² = Σ[(oi − ei)²/ei] Step 4: Computations with observed and expected frequencies

📌 Example: From the binomial data with n=5 and p=0.398 (calculated from sample mean x̄=1.99):

🔑 Definition — Chi-Square Test of Goodness of Fit: A hypothesis test that compares observed frequencies with expected frequencies from a theoretical distribution to determine if the theoretical distribution fits the observed data well.

Important Rule: Expected frequency for any category must not be less than 5. Categories with expected frequencies less than 5 must be merged with adjacent categories. Here, x=5 (ei=2.0) was merged with x=4 (ei=15.1).

Degrees of freedom = k − 1 − r = 5 − 1 − 1 = 3, where k=5 (merged categories) and r=1 (one parameter p estimated).

The critical value χ²0.05(3) = 7.82. Since computed χ² = 2.69 < 7.82, we accept H0 and conclude the binomial distribution fits the data well.

⭐ Key Takeaways

The RCB design ANOVA partitions total variation into treatment, block, and error components, allowing simultaneous testing of treatment effects and block effects with specific degrees of freedom formulas. The LSD test provides a single threshold value for comparing all pairs of treatment means after a significant ANOVA result, with means differing by more than the LSD considered significantly different. The Chi-Square Goodness of Fit test uses the statistic χ² = Σ[(o−e)²/e] with (k−1−r) degrees of freedom, where categories must be merged if expected frequencies are less than 5. A computed χ² less than the critical value indicates a good fit, meaning the theoretical distribution adequately represents the observed data. Blocking is beneficial only when block means are significantly different; otherwise, a Completely Randomized Design is preferred as it preserves more error degrees of freedom.

🧠 Quick Revision Questions

1. What are the degrees of freedom for error in an RCB design with r blocks and c treatments?
2. What is the formula for the Least Significant Difference, and what does each symbol represent?
3. Why must expected frequencies in a chi-square goodness of fit test not be less than 5?
4. In the chi-square goodness of fit test, how are the degrees of freedom calculated when parameters are estimated from the data?
5. What conclusion should be drawn if the computed F-value for blocks in an RCB design is less than the critical F-value?

📘 Lecture 45 — Chi-Square Test of Goodness of Fit (in continuation of the last lecture) • Chi-Square Test of Independence • The Concept of Degrees of Freedom • p-value • Relationship Between Confidence; Interval and Tests of Hypothesis An Overview of the Science of Statistics in Today’s World (including Latest Definition of Statistics)

📖 Overview: This lecture continues the discussion of the Chi-Square Test of Goodness of Fit, applying it to a Poisson distribution example. It then introduces the Chi-Square Test of Independence for analyzing associations between two categorical variables. The lecture also explains the fundamental concept of degrees of freedom, introduces the p-value method for hypothesis testing, explores the relationship between confidence intervals and hypothesis tests, and concludes with an overview of modern statistics.

🗂️ Topics Covered

The lecture covers the chi-square test of goodness of fit applied to a Poisson distribution example, including the step-by-step hypothesis testing procedure and the rationale behind the test. It then introduces the chi-square test of independence using contingency tables and examples. The concept of degrees of freedom is explained through geometric and algebraic analogies, followed by the definition and application of the p-value method. The lecture concludes with the relationship between confidence intervals and hypothesis tests, and provides an overview and latest definition of statistics.

📝 Lecture Summary

Chi-Square Test of Goodness of Fit (Continuation)

The lecture continues from the previous lecture's example of fitting a binomial distribution, now applying the chi-square test of goodness of fit to determine if customer arrivals at an airline terminal follow a Poisson distribution. The example uses data on the frequency of arrivals per minute. The process involves estimating the Poisson parameter (mean) from the sample data, calculating expected Poisson probabilities, and then comparing observed and expected frequencies using the chi-square statistic.

🔑 Definition — Chi-Square Test of Goodness of Fit: A statistical test used to determine how well an observed frequency distribution fits an expected theoretical distribution.

📐 Formula: The test statistic is ( \chi^2 = \sum \frac{(o_i - e_i)^2}{e_i} ), where ( o_i ) is the observed frequency and ( e_i ) is the expected frequency for category ( i ). This statistic, if the null hypothesis is true, follows a chi-square distribution with ( k - 1 - r ) degrees of freedom, where ( k ) is the number of categories after merging and ( r ) is the number of parameters estimated from the sample.

📌 Example: Data on customer arrivals per minute was collected (n=400). The sample mean was calculated as ( \bar{x} = \frac{800}{400} = 2 ). Using this, Poisson probabilities were computed [e.g., ( P(x=0) = \frac{e^{-2} 2^0}{0!} = 0.1353 )]. Expected frequencies were found by multiplying probabilities by 400. Because some expected frequencies were less than 5, the last four categories (arrivals 6, 7, 8, 9 or more) were combined, resulting in ( k = 7 ) effective categories.

• Hypotheses: ( H_0 ): Arrivals are Poisson-distributed; ( H_1 ): The distribution is not Poisson.
• Level of Significance: ( \alpha = 0.05 )
• Degrees of Freedom: Since one parameter (the mean) was estimated from the sample, ( r = 1 ). So, df = ( k - 1 - r = 7 - 1 - 1 = 5 ).
• Critical Value: ( \chi^2_{0.05}(5) = 11.07 )
• Computed Statistic: The sum of ( (o-e)^2/e ) across all categories yielded ( \chi^2 = 78.88 ).
• Conclusion: Since 78.88 > 11.07, we reject ( H_0 ) and conclude that the distribution is likely not a Poisson distribution with parameter 2. The large chi-square value indicates a very poor fit.

💡 Why this matters: This procedure is fundamental for validating model assumptions in real-world data. A poor fit means the assumed probability model (e.g., Poisson) is not suitable for the data.

Assumptions of the Chi-Square Test of Goodness of Fit:

1. The total sample size (n) should be at least 50.
2. The expected frequency (( e_i )) in any category should not be less than 5. Categories with ( e_i < 5 ) must be combined with adjacent categories.
3. Observations in the sample (or frequencies of categories) must be independent.

Chi-Square Test of Independence

This test determines whether two categorical variables of classification are associated or independent. It uses a contingency table and a formula very similar to the goodness-of-fit test.

🔑 Definition — Chi-Square Test of Independence: A statistical test used to determine if there is a significant association between two categorical variables in a population.

📐 Formula: The test statistic is ( \chi^2 = \sum_i \sum_j \frac{(o_{ij} - e_{ij})^2}{e_{ij}} ). This statistic, if ( H_0 ) is true, follows a chi-square distribution with ( (r-1)(c-1) ) degrees of freedom, where ( r ) and ( c ) are the number of rows and columns in the contingency table.

🔑 Definition — Expected Frequency in a Contingency Table: The expected frequency for any cell is calculated as: ( e_{ij} = \frac{\text{(Row Total)} \times \text{(Column Total)}}{\text{Grand Total}} ).

📌 Example: A sample of 250 men and 250 women were asked if they want a PC. The observed frequencies were: Men-Want (120), Men-Don’t Want (130), Women-Want (80), Women-Don’t Want (170).

• Hypotheses: ( H_0 ): Gender and desire for PC are independent; ( H_1 ): They are not independent.
• Significance Level: ( \alpha = 0.05 )
• Degrees of Freedom: ( (r-1)(c-1) = (2-1)(2-1) = 1 )
• Computations:
  • Expected frequency for Men-Want: ( e_{11} = (200 \times 250) / 500 = 100 )
  • Expected frequency for Women-Want: ( e_{12} = (200 \times 250) / 500 = 100 )
  • Expected frequency for Men-Don’t Want: ( e_{21} = (300 \times 250) / 500 = 150 )
  • Expected frequency for Women-Don’t Want: ( e_{22} = (300 \times 250) / 500 = 150 )
• Computed Statistic: ( \chi^2 = 4.00 + 2.67 + 4.00 + 2.67 = 13.33 )
• Critical Region: ( \chi^2 \ge \chi^2_{0.05}(1) = 3.84 )
• Conclusion: Since 13.33 > 3.84, we reject ( H_0 ) and conclude that desire to own a PC and sex are associated. The data shows the proportion wanting a PC is higher among men.

📌 Second Example: A ( 4 \times 3 ) contingency table of marital status (Single, Married, Widowed, Divorced) and smoking pattern (Abstinence, Only at times, Regular Smoker) is presented to test for independence. Students are encouraged to work on this problem.

The Concept of Degrees of Freedom

Degrees of freedom (df) can be understood as the number of values in a calculation that are free to vary, given certain constraints.

• Geometric Analogy: A line segment fixed at one point in a plane has 1 degree of freedom (it can rotate). A line segment fixed at both ends has 0 degrees of freedom in the plane.
• Statistical Analogy: For a sample of 6 observations with a fixed sum of 20, if you freely choose the first 5 values, the 6th value is determined. Therefore, you have ( n - 1 = 5 ) degrees of freedom.

🔑 Definition — Degrees of Freedom in Sampling Distributions: The number of observations in the sample minus the number of population parameters that are estimated from the sample data.

• For t-distribution (( t = \frac{\bar{x} - \mu_0}{s / \sqrt{n}} )): df = ( n-1 ) (one parameter, ( \sigma ), is estimated).
• For F-distribution (( F = s_1^2 / s_2^2 )): df = ( (n_1 - 1, n_2 - 1) ) (the variance parameter for each population is estimated).
• For Chi-Square Test of Independence (( \chi^2 ) for ( r \times c ) contingency table): df = ( (r-1)(c-1) ). In a ( 2 \times 2 ) table, given the marginal and grand totals, only one cell frequency can be freely chosen, so df = 1.

p-value

The p-value is an alternative method for hypothesis testing that indicates the strength of evidence against the null hypothesis.

🔑 Definition — p-value: The probability of observing a sample value as extreme as, or more extreme than, the value observed, given that the null hypothesis is true.

Decision Rule: If the p-value is less than the level of significance ( \alpha ), we reject ( H_0 ). If the p-value is greater than ( \alpha ), we do not reject ( H_0 ).

📌 Example: In the two-tailed z-test comparing hourly wages of computer analysts and nurses, the calculated test statistic was ( z = 3.43 ). Using the standard normal table:

• ( P(Z > 3.43) = 0.5000 - 0.4996 = 0.0004 )
• p-value = ( P(Z < -3.43) + P(Z > 3.43) = 0.0004 + 0.0004 = 0.0008 ).
• Since the p-value (0.0008) is less than ( \alpha = 0.02 ), we reject ( H_0 ). The result is highly improbable if the null hypothesis were true.

Relationship Between Confidence Interval and Tests of Hypothesis

There is a direct duality between a confidence interval for a parameter ( \theta ) and a two-tailed test of hypothesis about ( \theta ).

🔑 Definition — Proposition: Let [L, U] be a ( 100(1 - \alpha)% ) confidence interval for a parameter ( \theta ).

• We accept the null hypothesis ( H_0: \theta = \theta_0 ) at the ( \alpha ) level of significance if ( \theta_0 ) falls inside the interval [L, U].
• We reject ( H_0: \theta = \theta_0 ) at the ( \alpha ) level of significance if ( \theta_0 ) falls outside the interval [L, U].
• The confidence interval [L, U] is the acceptance region. The region outside is the rejection (or critical) region.

An Overview of the Science of Statistics in Today’s World

The lecture concludes by noting that this course covered fundamental concepts, but the field of statistics is vast, including topics like Latin Square Design, inference for regression, and non-parametric statistics. Many of the numerical problems demonstrated can be solved quickly using statistical software packages like SPSS, SAS, Minitab, etc.

Latest Definition of Statistics: Statistics is a science of decision making for governing state affairs. It collects, analyzes, manages, monitors, interprets, evaluates, and validates information. Statistics is Information Science and Information Science is Statistics. It is an applicable science whose tools are applied to all sciences, including humanities and social sciences.

⭐ Key Takeaways

1. The chi-square test of goodness of fit compares observed frequencies to expected frequencies from a theoretical distribution (e.g., Poisson, Binomial), and requires merging categories so that all expected frequencies are at least 5.
2. The chi-square test of independence determines if two categorical variables are associated, using a contingency table and calculating expected frequencies by multiplying the appropriate marginal totals and dividing by the grand total.
3. Degrees of freedom are the number of values free to vary in a calculation. For the goodness of fit test, it's ( k-1-r ); for the test of independence, it's ( (r-1)(c-1) ).
4. The p-value is the probability of observing your data (or more extreme) if the null hypothesis is true. A small p-value (less than ( \alpha )) provides evidence against the null hypothesis.
5. There is a direct relationship between a two-tailed hypothesis test and a confidence interval: if the hypothesized value falls within the ( 100(1-\alpha)% ) confidence interval, you fail to reject the null hypothesis at the ( \alpha ) level of significance.

🧠 Quick Revision Questions

1. What is the condition for merging categories in a chi-square goodness of fit test?
2. How is the expected frequency for a cell in a contingency table calculated for the test of independence?
3. Define degrees of freedom in the context of a t-test or a chi-square test.
4. What is the decision rule for rejecting a null hypothesis using the p-value method at a significance level ( \alpha )?
5. If a 95% confidence interval for ( \mu ) is (10, 20), what would be the conclusion of a two-tailed test of ( H_0: \mu = 22 ) at ( \alpha = 0.05 )?