π Lecture 39 β PATTERNS OF PROBABILITY: BINOMIAL, POISSON AND NORMAL DISTRIBUTIONS PART 3
π Overview: This lecture continues the exploration of probability distributions, covering practical applications of expected value and decision-making tools. It introduces the Poisson distribution for modeling rare events and demonstrates how to use decision tables and decision trees to make optimal business choices under uncertainty.
ποΈ Topics Covered
The lecture covers CRITBINOM example review, expected value calculation for lottery decisions, construction and use of decision tables for inventory optimization, decision tree analysis for manufacturing investment under uncertainty, and a comprehensive introduction to the Poisson distribution including its characteristics, formula, probability tables, and multiple worked examples for absences, breakdowns, and quality control scenarios.
π Lecture Summary
CRITBINOM EXAMPLE
The example shown under CRITBINOM in Handout 38 is reviewed as foundational material for understanding binomial cumulative distributions.
EXPECTED VALUE EXAMPLE
A lottery has 100 Rs. Payout on average 20 turns. The question is whether it is worthwhile to buy the lottery if the ticket price is 10 Rs.
Expected win per turn = p(winning) Γ gain per win + p(losing) Γ loss if you lose = 1/20 Γ (100 β 10) + 19/20 Γ (-10) = 90/20 β 190/20 = 4.5 β 9.5 = β5 Rs.
So on an average you stand to lose 5 Rs.
π‘ Why this matters: This demonstrates that a negative expected value means the game is unfavorable β a fundamental concept for evaluating any gambling or investment opportunity.
DECISION TABLES
A bakery problem is presented with the following data:
| No. of Pies demanded | % Occasions |
|---|---|
| 25 | 10 |
| 30 | 20 |
| 35 | 25 |
| 40 | 20 |
| 45 | 15 |
| 50 | 10 |
Given:
- Price per pie = Rs. 15
- Refund on return = Rs. 5
- Sale price = Rs. 25
- Profit per pie = Rs. 25 β 15 = Rs. 10
- Loss on each return = Rs. 15 β 5 = Rs. 10
To solve this problem, a decision table is set up. The values in the first column are number of pies to be purchased. Figures in columns are the sale with % share of sale within brackets. If the number of pies bought is less than the number that can be sold, the number of pies sold remains constant. If the number of pies bought exceeds the number of pies sold then the remaining are returned, meaning a loss.
Decision Table:
| Buy | 25(0.1) | 30(0.2) | 35(0.25) | 40(0.2) | 45(0.15) | 50(0.1) | EMV |
|---|---|---|---|---|---|---|---|
| 25 | 250 | 250 | 250 | 250 | 250 | 250 | 250 |
| 30 | 200 | 300 | 300 | 300 | 300 | 300 | 290 |
| 35 | 150 | 250 | 350 | 350 | 350 | 350 | 310 |
| 40 | 100 | 200 | 300 | 400 | 400 | 400 | 305 |
| 45 | 50 | 150 | 250 | 350 | 450 | 450 | 280 |
| 50 | 0 | 100 | 200 | 300 | 400 | 500 | 240 |
Expected profit for 30 pies: = 0.1 Γ 200 + 0.2 Γ 300 + 0.25 Γ 300 + 0.2 Γ 300 + 0.15 Γ 300 + 0.1 Γ 300 = 20 + 60 + 75 + 60 + 45 + 30 = 290 Rs.
Best Profit: The best profit is for 35 Pies = Rs. 310
DECISION TREE TOY MANUFACTURING CASE
The problem of a manufacturer intending to start manufacturing a new toy under conditions that a TV series may or may not appear, and that a rival may or may not sell a similar toy.
Decision Tree Structure: 1A β Abandon 1B β Go ahead β 2A: Series appears (60%) β 3A: Rival markets (50%) β 3B: No Rival (50%) β 2B: No series (40%)
Given data:
- Production: Series, no rival = 12000 units; Series, rival = 8000 units; No series = 2000 units
- Investment = Rs. 500,000
- Profit per unit = Rs. 200
- Loss if abandon = Rs. 500,000
Calculations:
- Profit if rival markets, series appears = 8000 Γ 200 β 500000 = 1,600,000 β 500,000 = 1,100,000 Rs.
- Profit if no rivals = 12000 Γ 200 β 500000 = 2,400,000 β 500,000 = 1,900,000 Rs.
- Profit/Loss if no series = 2000 Γ 200 β 500000 = 400,000 β 500,000 = β100,000 Rs.
EMV calculations:
- EMV (Series) = Rival markets and no rivals = 0.5 Γ 1,100,000 + 0.5 Γ 1,900,000 = 1,500,000
- EMV (Overall) = 0.6 Γ 1,500,000 + 0.4 Γ (β100,000) = 900,000 β 40,000 = 860,000 Rs.
Conclusion: It is clear that in spite of the uncertainty, there is a likelihood of a reasonable profit. Hence the conclusion is: Go ahead
THE POISSON DISTRIBUTION
The Poisson distribution is most commonly used to model the number of random occurrences of some phenomenon in a specified unit of space or time.
Examples:
- The number of phone calls received by a telephone operator in a 10-minute period
- The number of flaws in a bolt of fabric
- The number of typos per page made by a secretary
Characteristics:
- Either/or situation
- No data on trials
- No data on successes
- Average or mean value of successes or failures known
- Mean number of successes per unit, m, known and fixed
- p, chance, unknown but small (event is unusual)
π Definition β Poisson Distribution: A discrete probability distribution that expresses the probability of a given number of events occurring in a fixed interval of time or space if these events occur with a known constant mean rate and independently of the time since the last event.
π Formula: $$P(X = x) = \frac{e^{-\mu} \mu^x}{x!}, \quad x = 0,1,2,...$$
where ΞΌ is the average number of occurrences in the specified interval.
For the Poisson distribution: E(X) = ΞΌ, Var(X) = ΞΌ
π Example 1: The number of false fire alarms in a suburb of Houston averages 2.1 per day. The probability that 4 false alarms will occur on a given day is:
$$P(X = 4) = \frac{e^{-2.1} \times 2.1^4}{4!} = 0.0992$$
THE POISSON TABLES OF PROBABILITIES
Poisson tables give cumulative probability of r or more successes. Knowledge of m is required. The table gives the probability that r or more random events are contained in an interval when the average number of events per interval is m.
π Example 2: Attendance in a factory shows 7 absences on average. What is the probability that on a given day there will be more than 8 people absent?
Method 1: P(X > 8) = 1 β P(X β€ 8) = 1 β [P(x=0) + P(x=1) + P(x=2) + P(x=3) + P(x=4) + P(x=5) + P(x=6) + P(x=7) + P(x=8)]
= 1 β [0.0009 + 0.0064 + 0.0223 + 0.0521 + 0.0912 + 0.1277 + 0.149 + 0.149 + 0.1304] = 1 β 0.7291 = 0.2709
Method 2: P(X > 8) = 1 β P(X β€ 8) = 1 β 0.7291 = 0.2709
p(9 or more successes) = 0.2709
π Example 3: An automatic production line breaks down every 2 hours. Special production requires uninterrupted operation for 8 hours. What is the probability that this can be achieved?
ΞΌ = 8/2 = 4 x = 0 (no breakdown)
$$P(x=0) = \frac{e^{-4} \times 4^0}{0!} = 0.0183 = 1.83%$$
From Table: 1.83%
π Example 4: An automatic packing machine produces on average one in 100 underweight bags. What is the probability that 500 bags contain less than three underweight bags?
m = 1 Γ 500/100 = 5
p(x < 3, ΞΌ = 5) = p(x β€ 2, ΞΌ = 5) = 0.1247 = 12.47%
β Key Takeaways
The most critical points from this lecture are: Expected value is the fundamental concept for evaluating any probabilistic business decision β negative expected value means the decision is unfavorable. Decision tables allow comparing multiple purchasing/production quantities by calculating the Expected Monetary Value (EMV) for each option using weighted averages of possible outcomes. Decision trees model sequential decisions under uncertainty by applying the AND rule (multiplying probabilities along branches) and calculating EMV at each chance node. The Poisson distribution is essential for modeling rare events where we know the average rate (ΞΌ) but not the number of trials, with the key properties that E(X) = ΞΌ and Var(X) = ΞΌ. Poisson tables provide cumulative probabilities for "r or more" successes, saving calculation time for complex problems.
π§ Quick Revision Questions
- What is the expected value of the lottery with 100 Rs. payout on average every 20 turns and ticket price of 10 Rs.?
- In the pie decision table, what is the optimal number of pies to purchase and what is the resulting EMV?
- What are the four key characteristics of a Poisson distribution situation?
- In the factory absence example with ΞΌ=7, what is the probability of more than 8 absences on a given day?
- For an automatic packing machine with 1% defect rate, what is the probability that 500 bags contain less than 3 underweight bags?
π Lecture 40 β PATTERNS OF PROBABILITY: BINOMIAL, POISSON AND NORMAL DISTRIBUTIONS PART 4
π Overview: This lecture concludes the series on probability distributions by reviewing the Poisson distribution in Excel and introducing the Normal distribution for continuous data. It explains how to use the normal curve to find probabilities for real-world measurements (like candy bag weights) by calculating z-scores and using standard normal tables.
ποΈ Topics Covered
The lecture begins with the Excel POISSON worksheet functionβs syntax and examples for cumulative and exact probabilities. It then contrasts discrete distributions (Binomial/Poisson) with continuous distributions, using underweight candy boxes to introduce the normal distribution. Key topics include the standard normal distribution, reading areas under the normal curve from tables, calculating z-values for standardization, and solving four worked examples for proportions and percentages of bags above, below, or between specified weights.
π Lecture Summary
POISSON WORKSHEET FUNCTION
The Poisson distribution predicts the number of events over a specific time (e.g., cars arriving at a toll plaza in 1 minute). The Excel function POISSON(x, mean, cumulative) returns either the exact probability mass function or the cumulative probability.
π Definition β Poisson distribution: A discrete probability distribution that models the count of events occurring in a fixed interval of time or space, given a known average rate (mean).
π Formula:
For cumulative = FALSE (exact probability of exactly x events):
POISSON(x, mean, FALSE)
For cumulative = TRUE (probability of 0 to x events inclusive):
POISSON(x, mean, TRUE)
Remarks: x is truncated if not integer; if x or mean is nonnumeric, returns #VALUE!; if x β€ 0 or mean β€ 0, returns #NUM!.
π Example (Cumulative = TRUE): Returns the probability of βat most xβ events (e.g., at most 2 cars).
π Example (Cumulative = FALSE): Returns the probability of βexactly xβ events (e.g., exactly 2 cars).
THE PATTERN
In Binomial and Poisson distributions, situations are either/or, and the number of times can be counted β they are discrete probability distributions. In the Candy problem with underweight boxes, we measure weight β a continuous measurement β requiring a different treatment.
FREQUENCY BY WEIGHT
The frequency distribution of sample bag weights shows a distinct symmetrical shape, which is the shape of a Normal Distribution. A Standard Normal Distribution has mean = 0 and standard deviations 1, 2, 3, etc. marked on the x-axis.
NORMAL DISTRIBUTION
The blue curve is a typical Normal Distribution. A standard normal distribution has mean = 0 and standard deviation = 1. The Y-axis gives probability values; the X-axis gives z (measurement) values. Each point on the curve corresponds to the probability p that a measurement will yield a particular z value.
- Probability is a number from 0 to 1; percentage probabilities multiply p by 100.
- Area under the curve must be one.
- Probability is essentially zero for any value z greater than 3 standard deviations away from the mean on either side.
- Mean gives the peak of the curve; Standard deviation gives the spread.
π Weight distribution case: Mean = 510 g; StDev = 2.5 g.
Question: What proportion of bags weighs more than 515 g?
Answer: The proportion is given by the area under the curve to the right of 515 g.
π‘ Why this matters: The normal distribution is the foundation for many statistical tests and quality control applications in business and industry.
AREA UNDER THE STANDARD NORMAL CURVE
The normal distribution table gives the area under one tail only. The z-value ranges between 0 and 4 in the first column, and between 0 and 0.09 in the other columns.
π Example: Find area under one tail for z-value of 2.05.
- Look in column 1: find 2.0.
- Look in column 0.05: go to intersection of 2.0 and 0.05.
- Probability = 0.4798 (at the intersection, corresponding to z = 2.5).
Since probability from the center of the curve to the right is 0.5 and to the left is 0.5, to find probability for z β₯ 2.5, subtract the table value from 0.5: 0.5 β 0.4798 = 0.0202 = 2.02%
CALCULATING Z-VALUES
z = (Value x β Mean) / StDev
The process of calculating z from x is called Standardization. z indicates how many standard deviations the point is from the mean.
π Example 1: Find proportion of bags with weight in excess of 515 g.
Mean = 510 g, StDev = 2.5 g
z = (515 β 510) / 2.5 = 2
From tables: probability corresponding to z = 2 is 0.4772
Proportion = 0.5 β 0.4772 = 0.0228
π Example 2: What percentage of bags weigh less than 507.5 g?
Mean = 510 g, StDev = 2.5 g
z = (507.5 β 510) / 2.5 = β1
Look at z = +1: table probability = 0.3413
Percentage = 0.5 β 0.3413 = 0.1587 = 15.87%
π Example 3: Probability that a bag weighs less than 512 g?
z = (512 β 510) / 2.5 = 0.8
Table probability for z = 0.8 = 0.2881
Since 512 > mean, area = 0.5 + 0.2881 = 0.7881 = 78.81%
π Example 4: Percentage of bags weighing between 512 g and 515 g?
zβ = (512 β 510) / 2.5 = 0.8
Area to right of 512 = 0.5 β 0.2881 = 0.2119
zβ = (515 β 510) / 2.5 = 2
Area to right of 515 = 0.5 β 0.4772 = 0.0228
Required area = 0.2119 β 0.0228 = 0.1891 = 18.91%
β Key Takeaways
The Poisson distribution in Excel handles discrete event counts with a simple function, while the normal distribution applies to continuous measurements like weight. For the normal distribution, you must standardize raw measurements into z-scores using the formula z = (x β mean)/StDev. The standard normal table gives area under one tail (from the mean out to a given z); you add or subtract from 0.5 depending on whether your value is above or below the mean. For intervals between two values, compute the areas to the right of each z and subtract the larger from the smaller. These techniques are essential for making probability statements about any normally distributed business data.
π§ Quick Revision Questions
- What does the cumulative argument in the POISSON function control?
- How is the standard normal distribution defined?
- What is the total area under any normal probability distribution curve?
- How do you calculate the probability that a value is less than the mean using z-scores?
- In the candy bag example, what was the z-score for 515 g, and what proportion of bags weigh more than that?
π Lecture 41 β Estimating From Samples: Inference Part 1
π Overview: This lecture introduces the concept of statistical inference by examining how we estimate population characteristics from sample data. It covers the practical application of normal distribution functions in Excel (NORMDIST, NORMSDIST, NORMINV, NORMSINV) and establishes the foundation for understanding sampling distributions and standard error, which are critical for making reliable inferences about populations.
ποΈ Topics Covered
The lecture begins with a review of Excel functions for normal distribution calculations (NORMDIST, NORMSDIST, NORMINV, NORMSINV), then introduces the concept of sampling variations through practical examples involving defective components and household surveys. It explains how to construct a sampling distribution of percentages, calculates its mean, and introduces the Standard Error of Percentages (STEP) as a measure of sampling error. The lecture concludes that for large samples (n>30), the sampling distribution approximates a normal distribution.
π Lecture Summary
NORMDIST
This Excel function returns the normal distribution for a specified mean and standard deviation. It is widely used in statistics, including hypothesis testing.
Syntax: NORMDIST(x, mean, standard_dev, cumulative)
- X: The value for which you want the distribution
- Mean: Arithmetic mean of the distribution
- Standard_dev: Standard deviation of the distribution
- Cumulative: Logical value β TRUE returns cumulative distribution function; FALSE returns probability mass function
Remarks:
- If mean or standard_dev is nonnumeric, NORMDIST returns the #VALUE! error
- If standard_dev β€ 0, NORMDIST returns the #NUM! error
- If mean = 0, standard_dev = 1, and cumulative = TRUE, NORMDIST returns the standard normal distribution (NORMSDIST)
π Definition β Equation for normal density function (cumulative = FALSE): The formula is the standard normal density equation (not explicitly written in text, but referenced). When cumulative = TRUE, the formula is the integral from negative infinity to x of that equation.
π Example: x = 42, Arithmetic mean = 40, Standard deviation = 1.5, Cumulative distribution = 0.9
NORMSDIST
Returns the standard normal cumulative distribution function. The distribution has a mean of 0 and a standard deviation of 1. Use this function instead of a table of standard normal curve areas.
Syntax: NORMSDIST(z)
- z: The value for which you want the distribution
Remarks: If z is nonnumeric, NORMSDIST returns the #VALUE! error.
π Definition β Equation for standard normal density function: The equation is the standard normal density function formula (referenced in text).
π Example: Input z = 1.333333, Output cumulative probability = 0.908789
NORMINV
Returns the inverse of the normal cumulative distribution for a specified mean and standard deviation.
Syntax: NORMINV(probability, mean, standard_dev)
- Probability: A probability corresponding to the normal distribution
- Mean: Arithmetic mean of the distribution
- Standard_dev: Standard deviation of the distribution
Remarks:
- If any argument is nonnumeric, NORMINV returns #VALUE! error
- If probability < 0 or > 1, returns #NUM! error
- If standard_dev β€ 0, returns #NUM! error
- If mean = 0 and standard_dev = 1, uses standard normal distribution (NORMSINV)
- Uses iterative technique, accurate to within Β± 3x10^-7; if no convergence after 100 iterations, returns #N/A error
π Example: Given probability value, arithmetic mean, and standard deviation, the answer is the x-value.
NORMSINV
Returns the inverse of the standard normal cumulative distribution. The distribution has a mean of 0 and standard deviation of 1.
Syntax: NORMSINV(probability)
- Probability: A probability corresponding to the normal distribution
Remarks:
- If probability is nonnumeric, returns #VALUE! error
- If probability < 0 or > 1, returns #NUM! error
- Uses iterative technique, accurate to within Β± 3x10^-7; if no convergence after 100 iterations, returns #N/A error
π Example: Input is the z-value, corresponding cumulative distribution is calculated.
SAMPLING VARIATIONS
Electronic components are dispatched in boxes of 500. Customers agreed to a defect rate of 2%. One customer found 25 faulty components (5%) in a box. The question is whether this box was representative of production as a whole. The box represents a sample from the whole output. Sampling variations are expected. The key question: If the overall proportion of defective items has not increased, how likely is it that a box of 500 with 25 defective components will occur?
SAMPLING VARIATIONS EXAMPLE 1
In a residential colony section, there are 6 households (A, B, C, D, E, F). A survey determines the percentage of households that use corn flakes (cf) for breakfast. Households A, B, C, and D use corn flakes; E and F do not. Random samples of 3 households are taken.
All possible samples of size 3 and their percentage of cf users:
| Sample | % cf users | Sample | % cf users |
|---|---|---|---|
| ABC | 100 | BCD | 100 |
| ABD | 100 | BCE | 67 |
| ABE | 67 | BCF | 67 |
| ABF | 67 | BDE | 67 |
| ACD | 100 | BDF | 67 |
| ACE | 67 | BEF | 33 |
| ACF | 67 | CDE | 67 |
| ADE | 67 | CDF | 67 |
| ADF | 67 | CEF | 33 |
| AEF | 33 | DEF | 33 |
Out of 20 samples:
- 4 contain 100% cf users
- 12 contain 67% cf users
- 4 contain 33% cf users
Probability distribution:
- Probability of 100% cf users: 4/20 = 0.2
- Probability of 67% cf users: 12/20 = 0.6
- Probability of 33% cf users: 4/20 = 0.2
This is called a Sampling Distribution.
π‘ Why this matters: This illustrates that even when the true population percentage is fixed, different samples produce different percentages. This variability is natural and quantifiable.
SAMPLING DISTRIBUTION
The sampling distribution of percentages is the distribution obtained by taking all possible samples of fixed size n from a population, noting the percentage in each sample with a certain characteristic, and classifying these into percentages.
Mean of the Sampling Distribution Using the above data: Mean = (100% Γ 0.2) + (67% Γ 0.6) + (33% Γ 0.2) = 67%
The mean of the sampling distribution is the true percentage for the population as a whole. You must make allowance for variability in samples.
Conditions for Sample Selection:
- Number of items in the sample, n, is fixed and known in advance
- Each item either has or has not the desired characteristic
- The probability of selecting an item with the characteristic remains constant and is known to be P percent
If n is large (>30), then the distribution can be approximated to a normal distribution.
STANDARD ERROR OF PERCENTAGES
The standard deviation of the sampling distribution tells us how the sample values differ from the mean P. It gives us an idea of the error we might make if we were to use a sample value instead of the population value. For this reason, it is called the Standard Error of Percentages or STEP.
π Definition β STEP: The sampling distribution of percentages in samples of n items (n>30) taken at random from an infinite population in which P percent of items have characteristic X will be:
- A Normal Distribution
- With mean P%
- And standard deviation (STEP) = β[P(100-P)/n] %
The mean and standard deviation of the sampling distribution of percentages will also be percentages.
β Key Takeaways
The most critical concepts from this lecture are: (1) NORMDIST, NORMSDIST, NORMINV, and NORMSINV are Excel functions for working with normal distributions β the first two find probabilities from x or z values, while the last two find x or z values from probabilities; (2) sampling variation is natural and expected β different samples from the same population will yield different percentages; (3) the sampling distribution of percentages is constructed by taking all possible samples of a fixed size and recording the percentage with a characteristic; (4) the mean of the sampling distribution equals the true population percentage, making it an unbiased estimator; (5) the Standard Error of Percentages (STEP) = β[P(100-P)/n] % measures the typical error when using a sample percentage to estimate the population percentage, and for n>30, the sampling distribution approximates a normal distribution.
π§ Quick Revision Questions
- What is the difference between NORMDIST and NORMINV in terms of what they calculate?
- In the household survey example with 6 households, what was the mean of the sampling distribution and what does it represent?
- What are the three conditions required for sample selection when constructing a sampling distribution of percentages?
- What does the Standard Error of Percentages (STEP) measure, and why is it important for inference?
- Under what condition can the sampling distribution of percentages be approximated by a normal distribution, and what are its parameters (mean and standard deviation)?
π Lecture 42 β Estimating from Samples: Inference Part 2
π Overview: This lecture continues the study of statistical inference by focusing on estimating population parameters from sample data. It covers how to calculate the probability of sample outcomes, construct confidence intervals for population percentages, determine required sample sizes, and understand the distribution of sample means, providing essential tools for making data-driven decisions.
ποΈ Topics Covered
The lecture begins with a review example calculating the probability of a sample containing a certain number of women. It then introduces the applications of the Standard Error of Percentage (STEP), explains the process of calculating 95% and 99% confidence limits for population percentages, and shows how to determine the necessary sample size for a desired accuracy. Finally, it covers the distribution of sample means, introducing the Standard Error of the Mean (STEM) and an example probability calculation.
π Lecture Summary
Review of Lecture 41
A factory has 25% women workers. We need to find the probability that a random sample of 80 workers contains 25 or more women.
The Standard Error of Percentage (STEP) is calculated as the standard deviation of the sampling distribution of percentages. Its formula is used to find the z-score for the sample percentage.
- Mean = P = 25%
- n = 80
- STEP = [P(100-P)/n]^(1/2) % = [25(75)/80]^(1/2) % = 4.84%
- % women in sample = (25/80) x 100% = 31.25%
- z = (31.25 - 25) / 4.84 = 1.29
Looking up z = 1.29 in the normal distribution table gives a probability of 0.4015 (the area between the mean and z). Since we want the probability of 25 or more women (the area to the right of z), we calculate:
p(sample contains 25 or more women) = 0.5 - 0.4015 = 0.0985 or about 10%.
π Example: In a factory, 25% of the workforce is women. The probability that a random sample of 80 workers contains 25 or more women is approximately 10%.
Applications of STEP
Important issues addressed by using STEP include:
- Determining the probability that a specific sample will arise.
- Estimating the population percentage P from the information obtained from a single sample.
- Calculating the required sample size to estimate a population percentage with a given degree of accuracy.
Confidence Limits
A market researcher selects a random sample of 400 consumers. 280 (70%) are purchasers of the companyβs product. The goal is to conclude the percentage of all consumers buying the product.
The 95% confidence limits are commonly used. In a normal sampling distribution, 2.5% (rejection region) corresponds to a z-value of 1.96 on either side of the sample percentage (70%). The acceptance region is 0.5 - 0.025 = 0.475, for which z = 1.96.
The sample percentage (70%) is used as an approximation for the population percentage P.
- STEP = [70(100-70)/400]^1/2 = 2.29%
- Confidence Limits: Estimate for population percentage = 70 +/- 1.96 Γ STEP
- = 70 +/- 1.96 Γ 2.29
- = 65.515% and 74.49%
- A common approximation is to round 1.96 to 2. Therefore, the 95% confidence interval is P +/- 2 Γ STEP.
π Example: A sample of 60 students contains 12 (20%) left-handed students. The 95% confidence interval for the percentage of all left-handed students is: Range = 20 +/- 2 Γ STEP Range = 20 +/- 2 Γ [20 Γ (100-20) / 60]^1/2 = 9.67% and 30.33%
Estimating Process Summary:
- Identify the sample size n and the sample percentage P.
- Calculate STEP using these values.
- The 95% confidence interval is approximately P +/- 2 Γ STEP.
99% Confidence
For 99% confidence limits, the z-value is 2.58. At a 99% confidence level, there is a 1% chance of error (level of significance), with 0.5% rejection region on each side of the curve. The acceptance region is 0.5 - 0.005 = 0.495, for which z = 2.58.
Finding a Sample Size
To satisfy 95% confidence, we can set a desired margin of error. For example, to ensure the estimate is within 5% of the true answer:
- 2 Γ STEP = 5
- STEP = 2.5
- Assuming a pilot survey value of P = 30%:
- STEP = [30 Γ 70 / n]^1/2 = 2.5
- Solving for n gives n = 336.
π Example: To be 95% confident that an estimate of a population percentage is within 5% of the true value, a sample of 336 persons must be interviewed.
Distribution of Sample Means
The standard deviation of the Sampling Distribution of means is called the Standard Error of the Mean (STEM).
π Formula: STEM = s.d / βn
- s.d denotes the standard deviation of the population.
- n is the size of the sample.
π Example: What is the probability that a random sample of 64 children from a population with mean IQ = 100 and StDev = 15 will have a sample mean IQ below 95?
- s = 15; n = 64; population mean = 100
- STEM = 15 / β64 = 15 / 8 = 1.875
- z = (100 - 95) / 1.875 = 2.67
- This gives a probability of 0.0038. The chance that the average IQ of the sample is below 95 is very small.
β Key Takeaways
- The Standard Error of Percentage (STEP) quantifies the variability of sample percentages and is used to calculate z-scores for probabilities and confidence intervals for population percentages.
- The 95% confidence interval for a population percentage is approximately P Β± 2 Γ STEP, where P is the sample percentage. This is a core method for estimation.
- To achieve a desired margin of error (e.g., 5%) with 95% confidence, you can calculate the required sample size using the formula derived from the confidence interval equation.
- The Standard Error of the Mean (STEM) measures the variability of sample means and is calculated as the population standard deviation divided by the square root of the sample size.
- Understanding the difference between STEP (for percentages/proportions) and STEM (for means) is critical, as each is used in different types of inference problems.
π§ Quick Revision Questions
- What is the formula for the Standard Error of Percentage (STEP)?
- What z-value is used for a 99% confidence interval?
- If a sample of 400 consumers shows 70% are purchasers, calculate the 95% confidence interval for the population percentage.
- What is the formula for the Standard Error of the Mean (STEM)?
- A sample mean is 95, while the population mean is 100. The STEM is 1.875. Calculate the z-score for the sample mean.
π Lecture 43 β Hypothesis Testing: Chi-Square Distribution Part 1
π Overview: This lecture introduces hypothesis testing using the normal approximation to the binomial distribution. It explains how to test whether a sample differs significantly from a known population proportion, using the Null Hypothesis framework. The lecture also covers the Finite Population Correction Factor and reviews sampling error concepts.
ποΈ Topics Covered
The lecture begins with a review of sampling error using an example of tins of beans, then revisits the problem of faulty components. It introduces the Finite Population Correction Factor and presents the Training Manager's Problem as a case study for hypothesis testing. The core concepts of Null Hypothesis, STEP calculation, confidence intervals, and the procedure for carrying out hypothesis tests are explained with worked examples. The lecture concludes with further points about error types in hypothesis testing.
π Lecture Summary
Review Lecture 42
Example 1 β Confidence Interval for Population Mean An inspector took a sample of 100 tins of beans. The sample weight is 225 g. Standard deviation is 5 g. Calculate with 95% confidence the range of the population mean.
Since the population standard deviation is not known, use the sample standard deviation as an approximation.
π Formula: STEM = s.d / βn = 5 / β100 = 5 / 10 = 0.5
π Example: 95% confidence interval = 225 Β± 2 Γ 0.5 = from 224 to 226 g.
Problem of Faulty Components Revisited
A box of 500 components may have 25 or 5% faulty components. Overall faulty items = 2%. P = 2%; n = 500
π Formula: STEP = β[P(100 β P)/n] = β[(2 Γ 98)/500] = β0.392 = 0.626
To find the probability that the sample percentage is 5% or over: z = (5 β 2)/STEP = 3/0.626 = 4.79
Area against z = 4.79 is negligible. The chance of such a sample is very small.
Finite Population Correction Factor
If the population is very large compared to the sample, multiply STEM and STEP by the:
π Formula: Finite Population Correction Factor = β[1 β (n/N)]
Where:
- N = Size of the population
- n = Size of the sample
- n must be less than 0.1N (sample is less than 10% of population)
π Definition β Finite Population Correction Factor: A multiplier applied to standard errors when the sample size exceeds 10% of the population size, correcting for the reduced variability when sampling without replacement from a finite population.
π‘ Why this matters: Without this correction, you would overestimate the standard error for large samples relative to population size, leading to wider confidence intervals than necessary.
Training Manager's Problem
A new refresher course for training of workers was completed. The Training Manager would like to assess the effect of retraining if any.
Particular questions:
- Is quality of product better than produced before retraining?
- Has the speed of machines increased?
- Do some classes of workers respond better to retraining than others?
The Training Manager hopes to:
- Compare the new position with established standard deviation
- Test a theory or hypothesis about the course
Case Study Before the course: Worker X produced 4% rejects. After the course: Out of 400 items, 14 were defective = 3.5%. An improvement?
The 3.5% figure may not demonstrate overall improvement. It does not follow that every single sample of 400 items contains exactly 4% rejects. To draw a sound conclusion, sampling variations must be taken into account.
We do not begin by assuming what we are trying to prove β we use the Null Hypothesis. We must begin with the assumption that there is no change at all.
π Definition β Null Hypothesis: The initial assumption that there is no change or no difference; that the observed results are due to chance or sampling variation alone. It is the hypothesis we try to disprove.
Implication of Null Hypothesis: That the sample of 400 items taken after the course was drawn from a population in which the percentage of reject items is still 4%.
Null Hypothesis Example Data: P = 4%; n = 400
π Formula: STEP = β[P(100 β P)/n] = β[4(100 β 4)/400] = β[4 Γ 96/400] = β0.96 = 0.98%
At 95% confidence limit: Range = 4 Β± 2 Γ 0.98 = 2.04 to 5.96%
π Example: The sample with 3.5% rejects falls inside the interval 2.04% to 5.96%. Conclusion: The sample is not inconsistent with the Null Hypothesis. There were no grounds for rejecting the Null Hypothesis.
Another Example Before the course: 5% rejects After the course: 2.5% rejects (10 out of 400)
P = 5 STEP = β[5(100 β 5)/400] = β[5 Γ 95/400] = β1.1875 = 1.09
Range at 95% Confidence Limits: = 5 Β± 2 Γ 1.09 = 2.82% to 7.18%
π Example: The sample with 2.5% rejects falls outside the interval 2.82% to 7.18%. Conclusion: Doubt about Null Hypothesis most of the time. The Null Hypothesis should be rejected.
Procedure for Carrying Out Hypothesis Test
- Formulate the Null Hypothesis
- Calculate STEP and P Β± 2 Γ STEP (the 95% confidence interval)
- Compare the sample % with this interval to see whether it is inside or outside
Decision Rule:
- If the sample falls outside the interval: reject the Null Hypothesis (sample differs significantly from the population %)
- If the sample falls inside the interval: do not reject the Null Hypothesis (sample does not differ significantly from the population % at 5% level)
How the Rule Works
The bigger the difference between the sample and population percentages, the less likely it is that the population percentages will be applicable.
- When the difference is so big that the sample falls outside the 95% interval, then the population percentages cannot be applied. The Null Hypothesis must be rejected.
- If the sample belongs to the majority and it falls within the 95% interval, then there are no grounds for doubting the Null Hypothesis.
Further Points About Hypothesis Testing
- 99% interval requires 2.58 Γ STEP. The interval becomes wider. It is less likely to conclude that something is significant.
- Two types of errors can occur:
π Definition β Type 1 Error: We might conclude there is a significant difference when there is none. The chance of this error equals the significance level (5% when using 95% confidence interval).
π Definition β Type 2 Error: We might decide that there is no significant difference when there is one.
π‘ Why this matters: The choice between 95% and 99% confidence levels involves a trade-off between these two error types. A wider interval (99%) reduces Type 1 error but increases Type 2 error.
β Key Takeaways
The Null Hypothesis is the assumption of no change, and hypothesis testing determines whether sample evidence is strong enough to reject it. The test involves calculating STEP using the formula β[P(100βP)/n], constructing a 95% confidence interval around the population proportion (P Β± 2ΓSTEP), and checking whether the sample proportion falls inside or outside this interval. If the sample falls outside, we reject the Null Hypothesis (significant difference exists); if inside, we do not reject it. The Finite Population Correction Factor β[1β(n/N)] must be applied when the sample exceeds 10% of the population. Two types of errors exist: Type 1 (false positive β concluding significance when none exists, risk = 5% at 95% confidence) and Type 2 (false negative β missing a real difference).
π§ Quick Revision Questions
- What is the Null Hypothesis, and why must we begin a hypothesis test with this assumption?
- Calculate STEP for a population with 8% defective items and a sample of 200 items.
- If a sample of 300 items has 12% defectives and the population has 10% defectives, does this reject or fail to reject the Null Hypothesis at the 95% confidence level?
- When must the Finite Population Correction Factor be applied, and what is its formula?
- What is the difference between Type 1 Error and Type 2 Error in hypothesis testing?
π Lecture 44 β Hypothesis Testing: Chi-Square Distribution Part 2
π Overview: This lecture continues the study of hypothesis testing, building on concepts from Lecture 43. It covers detailed mechanics of one-tailed and two-tailed tests, introduces hypothesis testing for means and small samples using the t-distribution, and explains how to test differences between two sample means and more than one proportion using the Chi-Square distribution.
ποΈ Topics Covered
The lecture covers further points about hypothesis testing including one-tailed and two-tailed tests, hypothesis testing about means with worked examples, alternative hypothesis testing using z-values, a summary of the hypothesis testing process, testing hypotheses about small samples, the Studentβs t-distribution, a summary of when to use z-tests vs. t-tests, testing differences between two sample means, testing more than one proportion using Chi-Square, and the use of Excelβs CHITEST function.
π Lecture Summary
Further Points About Hypothesis Testing
This section continues from Handout 43 and explains that we cannot draw conclusions about the direction of a difference without specifying a tail. Two possibilities are demonstrated: a one-tailed test and a two-tailed test.
π Definition β One-tailed test: A test where the alternative hypothesis specifies a direction (greater than or less than). π Formula:
- Range (one-tailed 5%): ( \text{Population proportion} - 1.64 \times \text{STEP} )
- Range (two-tailed 5%): ( \text{Population proportion} \pm 1.96 \times \text{STEP} ) π Example (One-tailed): Null Hypothesis: ( P \ge 4% ) vs. ( P > 4% ). STEP = 0.98%. Range = ( 4 - 1.64 \times 0.98 = 2.39% ). New figure = 3.5%. Since 3.5% > 2.39%, there is no reason to conclude improvement. π Example (Two-tailed): Same null. Range = ( 4 \pm 1.96 \times 0.98 = 2.08% ) to 5.92%. New figure = 3.5%. Since 3.5% falls within the range, no reason to conclude change.
Hypotheses About Means
This revisits the retraining course problem. Before the course: worker took 2.5 minutes per item, Standard Deviation = 0.5 min. After the course: sample of 64 items, mean = 2.58 min. Null hypothesis: no change.
π Definition β STEM: Standard Error of the Mean = ( \frac{\text{population s.d.}}{\sqrt{n}} ) π Formula: STEM = ( 0.5 / \sqrt{64} = 0.0625 ) π Example: Range = ( 2.5 \pm 2 \times 0.0625 = 2.375 ) to 2.625 min. Sample mean (2.58) falls within this range. Conclusion: No grounds for rejecting the null hypothesis at 5% significance level.
Alternative Hypothesis Testing Using Z-Value
When testing proportion, we can compute a sample z-value and compare it with the critical z-value from the standard normal distribution.
π Formula: ( z = \frac{\text{sample percentage} - \text{population mean}}{\text{STEP}} ) π Example: ( z = (3.5 - 4)/0.98 = -0.51 ). Compare with critical z β 1.96 (5% two-tailed). Since 0.51 < 1.96, the probability of getting this sample by chance is greater than 5%. Sample is consistent with null hypothesis; do not reject.
Process Summary
This outlines the six-step hypothesis testing procedure:
- State Null Hypothesis (1-tailed or 2-tailed)
- Decide significance level and find critical value of z
- Calculate sample z = (sample value β population value) Γ· STEP or STEM
- Compare sample z with critical z
- If sample z is smaller, do not reject Null Hypothesis
- If sample z is greater, reject Null Hypothesis
Testing Hypotheses About Small Samples
Large sample means are normally distributed regardless of the underlying distribution, but this does not apply to small samples. For small samples, hypothesis testing requires that the underlying distribution is normal. If we only know the sample Standard Deviation and must approximate the population s.d., we use Studentβs t-distribution.
Studentβs t-Distribution
Studentβs t-distribution is very similar to the normal distribution but is wider, reflecting greater uncertainty. As sample size n increases, the t-distribution approximates the normal distribution.
π Definition β t-distribution: A family of distributions used when sample size is small and population standard deviation is unknown; depends on degrees of freedom (v = n β 1). π Formula: STEM = ( \frac{\text{sample s.d.}}{\sqrt{n}} ) (using n β 1 divisor for s.d. calculation) π Example: Population mean training time = 10 days. Sample of 8 women: mean = 9 days, sample s.d. = 2 days. STEM = ( 2/\sqrt{8} = 0.71 ). Null: no difference. t = ( (9 - 10)/0.71 = -1.41 ). v = 7. Critical t at 5% (2-tailed) = 2.365. Since 1.41 < 2.365, do not reject null hypothesis.
Summary - I
If underlying population is normal and we know the Standard Deviation, then distribution of sample means is normal with STEM = population s.d./βn, and we can use a z-test.
Summary - II
If underlying population is unknown but sample is large, then distribution of sample means is approximately normal with STEM = population s.d./βn, and we can use a z-test.
Summary - III
If underlying population is normal but we do not know its s.d. and sample is small, we use sample s.d. with n β 1 divisor. Distribution of sample means follows a t-distribution with n β 1 degrees of freedom, and we can use a t-test.
Summary - IV
If underlying population is not normal and we have a small sample, none of the hypothesis testing procedures can be safely used.
Testing Difference Between Two Sample Means
This section tests whether there is a significant difference between the mean wages of two groups: 30 production workers (mean = Rs. 120, s.d. = 10) and 50 maintenance workers (mean = Rs. 130, s.d. = 12).
π Definition β STEDM: Standard Error of Difference in Sample Means π Formula:
- ( s = \sqrt{\frac{n_1 s_1^2 + n_2 s_2^2}{n_1 + n_2}} )
- STEDM = ( s \times \sqrt{\frac{1}{n_1} + \frac{1}{n_2}} ) π‘ Why this matters: This allows testing whether two independent samples come from populations with the same mean. π Example: ( s = \sqrt{(30 \times 100 + 50 \times 144)/(80)} = 11.29 ). STEDM = ( 11.29 \times \sqrt{1/30 + 1/50} = 2.60 ). ( z = (120 - 130)/2.60 = -3.85 ). This is well outside the critical z for 5% significance. Null hypothesis (no difference) is rejected.
Procedure Summary
- State Null Hypothesis and decide significance level
- Identify information and decide standard error and distribution needed
- Calculate standard error
- Calculate z or t = difference Γ· standard error
- Compare with critical value; if greater, reject null hypothesis
More Than One Proportion
When analyzing improvement across age groups, we can test if the observed frequencies differ from expected frequencies under the assumption of uniform improvement.
π Definition β Chi-squared (ΟΒ²): A statistic measuring the disagreement between observed and expected frequencies: ( \chi^2 = \sum \frac{(O - E)^2}{E} ) π Formula: Degrees of freedom, v = (r β 1) Γ (c β 1), where r = rows, c = columns π Example: Data:
- Under 35: 17 improved (14 expected), 4 did not (7 expected)
- 35-50: 17 improved (16 expected), 7 did not (8 expected)
- Over 50: 6 improved (10 expected), 9 did not (5 expected) Calculations: O β E = [3, 1, -4, -3, -1, 4]; (O-E)Β² = [9, 1, 16, 9, 1, 16]; (O-E)Β²/E = [0.643, 0.0625, 1.6, 1.286, 0.125, 3.2]; Total ΟΒ² = 6.92. v = (3-1)(2-1) = 2. Critical ΟΒ² at 5% = 5.991. Since 6.92 > 5.991, null hypothesis (uniform improvement) is rejected.
Chi-Squared Summary
- Formulate null hypothesis (no association)
- Calculate expected frequencies
- Calculate ΟΒ²
- Calculate degrees of freedom = (rows β 1) Γ (columns β 1); find critical ΟΒ²
- Compare; if sample ΟΒ² is smaller, donβt reject null; if larger, reject null
CHITEST in Excel
The CHITEST function returns the test for independence. It returns the probability from the chi-squared distribution for the statistic and appropriate degrees of freedom.
π Definition β CHITEST: Returns ( P(X > \chi^2) ), where ( \chi^2 = \sum \frac{(A_{ij} - E_{ij})^2}{E_{ij}} ), with degrees of freedom df = (r β 1)(c β 1). π Example: For data with two groups, the probability for chi-squared = 16.16957 with 2 degrees of freedom was 0.000308, which is negligible, indicating strong evidence against the null hypothesis.
β Key Takeaways
The most critical concepts from this lecture are: the distinction between one-tailed and two-tailed tests and how to compute ranges and z-values accordingly; the correct selection of z-test versus t-test based on sample size and knowledge of population standard deviation; the formula and interpretation of the Studentβs t-distribution for small samples; the computation and meaning of the Standard Error of Difference in Means (STEDM) for comparing two samples; and the Chi-Square test for independence, including calculation of expected frequencies, ΟΒ² statistic, and degrees of freedom, along with the critical value comparison to determine whether to reject the null hypothesis.
π§ Quick Revision Questions
- What is the difference between a one-tailed and a two-tailed hypothesis test, and how does this affect the critical value used?
- Under what conditions should a t-test be used instead of a z-test for hypothesis testing about means?
- How is the Standard Error of Difference in Means (STEDM) calculated when comparing two independent samples?
- What is the formula for the Chi-Square statistic, and how are degrees of freedom determined for a contingency table?
- In a Chi-Square test, if the calculated ΟΒ² = 6.92 and the critical value at 5% with v = 2 is 5.991, what conclusion should be drawn?