MTH001 — Final Term Summary (Lectures 23–32)

📘 Lecture 23 — Data Representation

📖 Overview: This lecture introduces the fundamental techniques for summarizing and describing qualitative data, both in univariate and bivariate situations. It covers tabulation methods and various graphical representations, explaining how to choose the appropriate chart based on the nature of the data. Understanding these methods is crucial for effectively communicating data insights without losing essential information.

🗂️ Topics Covered

The lecture begins by outlining two broad categories of data: qualitative and quantitative. It then focuses on qualitative data, starting with univariate frequency tables and the pie chart. Next, it covers the simple bar chart for time-series or ordered data. The discussion expands to bivariate frequency tables for two qualitative variables, followed by the component bar chart for comparing totals and their parts, and finally the multiple bar chart for comparing two distinct but related variables.

📝 Lecture Summary

Data Representation

The lecture begins by reiterating the two main categories of data: qualitative and quantitative. For summarizing and describing these types of data, a variety of methods exist, as outlined in a tree-diagram.

🔑 Definition — Qualitative Data: Data that can be classified into non-numerical categories, such as "Urdu Medium" or "English Medium". 🔑 Definition — Quantitative Data: Numerical data that can be either discrete or continuous.

💡 Why this matters: The first step in any data analysis is correctly identifying the data type, as this determines which summarization and visualization techniques are valid and appropriate to use.

Univariate Frequency Table

Summarizing a single qualitative variable begins by counting the number of observations in each category. This creates a univariate frequency table.

📌 Example: In a survey of 1200 first-year students about their medium of schooling (Urdu or English), the results are:

• 719 students from Urdu medium schools
• 481 students from English medium schools

This raw count is the frequency (f). To make the data more useful, proportions or percentages are calculated.

• Urdu: 719 / 1200 = 0.599 ≈ 60%
• English: 481 / 1200 = 0.401 ≈ 40%

Pie Chart

A pie chart is a circular diagram divided into sectors, where the size of each sector is proportional to its category's frequency. The angle for each sector is determined by dividing the cell frequency by the total frequency and multiplying by 360°.

📐 Formula: Sector Angle = (Category Frequency / Total Frequency) × 360°

📌 Example: For the student survey:

• Urdu Sector Angle: (719 / 1200) × 360° = 215.7°
• English Sector Angle: (481 / 1200) × 360° = 144.3°

This creates a visual representation where the circle is cut into two sectors representing the two mediums of schooling.

Simple Bar Chart

A simple bar chart consists of horizontal or vertical bars of equal width, with lengths proportional to the values they represent. It is useful for comparing distinct categories or values over time.

🔑 Definition — Simple Bar Chart: A chart with bars of equal width where the length of each bar is proportional to the value it represents. When values do not relate to time, they should be arranged in ascending or descending order before charting.

📌 Example: The turnover of a company for the years 1965 to 1969 is shown below:

In the bar chart, the x-axis represents the year, and the y-axis has a scale for turnover. A vertical bar is drawn against each year, with a height corresponding to the turnover figure.

Bivariate Frequency Table

When summarizing two qualitative variables simultaneously, a bivariate frequency table is constructed. This table has a "boxhead" (top row) and a "stub" (first column).

🔑 Definition — Bivariate Frequency Table: A table used to summarize the frequencies of two or more categorical variables simultaneously.

📌 Example: Revisiting the student survey, if we also record the sex of each student, we get a bivariate situation. The table categorizes students into four groups: Male/Urdu, Female/Urdu, Male/English, and Female/English.

Component Bar Chart

A component bar chart (or subdivided bar chart) is used to display the totals of different groups and their components. Each bar represents a total, and is subdivided into parts representing the components that make up that total.

🔑 Definition — Component Bar Chart: A bar chart where each bar is divided into parts, usually stacked on top of each other, to show how a whole is broken down into its constituent components.

📌 Example: For the bivariate student data, two bars are drawn: one for "Male" and one for "Female". Each bar is subdivided into two parts:

• The lower part represents students from English medium schools.
• The upper part represents students from Urdu medium schools. This allows for a quick comparison of both the total number of male vs. female students and the composition of their schooling medium.

💡 Why this matters: The component bar chart is ideal when the components (e.g., Urdu and English students) logically add up to the total for each group (e.g., total male students).

Multiple Bar Chart

A multiple bar chart consists of sets of grouped bars, where each bar in a group represents a different variable. The bars are shaded or coloured differently to aid identification.

🔑 Definition — Multiple Bar Chart: A bar chart that uses two or more sets of bars to compare two or more distinct variables across the same categories (e.g., years).

📌 Example: Comparing imports and exports of Pakistan from 1970-71 to 1974-75.

In the chart, for each year, one bar represents the value of imports and an adjacent bar represents the value of exports. This provides a clear visual comparison of the two variables over time. The basic difference is that in a component bar chart, the parts add up to a total, whereas in a multiple bar chart, they do not.

💡 Why this matters: The multiple bar chart is excellent for comparing two or more separate but related trends, like imports vs. exports, or sales of product A vs. product B.

⭐ Key Takeaways

The most critical exam points from this lecture are the distinction between a pie chart and a bar chart for univariate data, and the crucial difference between a component bar chart and a multiple bar chart for bivariate data. You must know that a component bar chart is for data where categories add up to a total (like parts of a whole), whereas a multiple bar chart compares separate, non-additive variables (like imports and exports). Also, remember that a simple bar chart is used for a single variable, and its values should be ordered if they are not time-based. Finally, the sector angle formula for a pie chart is (frequency / total) * 360°.

🧠 Quick Revision Questions

1. What is the first step in summarizing a univariate qualitative dataset?
2. Explain the key difference between a component bar chart and a multiple bar chart, using an example for each.
3. In a pie chart, how is the sector angle for a category calculated?
4. When constructing a simple bar chart for data that does not relate to time, what arrangement should the values be in?
5. A survey records a student's "favorite subject" (Math, Science, English) and their "grade level" (9, 10, 11). Which type of chart would best represent this bivariate data?

📘 Lecture 24 — The Central Tendency of a Data Set

📖 Overview: This lecture continues the discussion on measures of central tendency, focusing on the mode, its properties, and special cases like non-modal and bi-modal distributions. It then introduces the arithmetic mean as the most widely used average, covering its calculation for both ungrouped and grouped data, along with the concept of grouping error. Finally, the lecture explains the weighted arithmetic mean and the median, illustrating their applications and advantages in various scenarios.

🗂️ Topics Covered

The lecture revisits the formula for calculating the mode in a grouped frequency distribution and its desirable properties. It then introduces and defines the arithmetic mean, demonstrating its computation for raw data and grouped data using the class-mark method, while discussing the grouping error. The concept of the weighted arithmetic mean is introduced with a practical example, followed by a detailed explanation of the median, including its calculation for raw data and discrete frequency distributions.

📝 Lecture Summary

The Mode

The lecture begins by revising the mode calculation from the previous lecture, using the example of EPA mileage ratings. The formula for the mode in a grouped frequency distribution is:

Mode:

X̂ = l + [(f_m - f_1) / (f_m - f_1) + (f_m - f_2)] * h

Where:

• l = lower class boundary of the modal class
• f_m = frequency of the modal class
• f_1 = frequency of the class preceding the modal class
• f_2 = frequency of the class following the modal class
• h = length of class interval of the modal class

For the EPA mileage data, the modal class is 36.0 – 38.9 with a frequency of 14. X̂ = 35.95 + [(14 - 4) / (14 - 4) + (14 - 8)] * 3 X̂ = 35.95 + [10 / 16] * 3 X̂ = 35.95 + 1.875 = 37.825

The mode lies in the middle of the frequency distribution, qualifying it as a measure of central tendency.

🔑 Definition — Mode: The value that occurs most frequently in a data set.

Desirable Properties of the Mode:

• The mode is easily understood and ascertained for discrete frequency distributions.
• It is not affected by a few very high or low values.

When to use the Mode: The mode is valuable for inventory problems. For example, a clothing store manager would use the modal hat size to stock the greatest quantity, as this is the size sold most often. 💡 Why this matters: The mode is a practical average for business decisions where the most common item is needed.

Special Cases:

• No Mode: A simple series may have no mode if no value occurs more than once.
• Bi-modal Distribution: A frequency distribution with two modes is called bi-modal.

The Arithmetic Mean

The arithmetic mean is the statistician's term for the average. It is the most widely used and easiest measure of central tendency to calculate.

🔑 Definition — Arithmetic Mean: A value obtained by dividing the sum of all observations by their number.

For Ungrouped Data: X̄ = ΣX / n

Formula: X̄ = (Sum of all observations) / (Number of observations)

📌 Example: A news agent's weekly receipts are: £9.90, £7.75, £19.50, £32.75, £63.75, £75.50, £50.70. X̄ = £259.85 / 7 = £37.12

Interpretation: The mean, £37.12, represents the amount that would have been obtained each day if the same amount were obtained every day.

For Grouped Data (Frequency Distribution): When data is grouped into a frequency distribution, the identity of individual observations is lost. Therefore, the observations in each class are assumed to be identical with the class midpoint (Xi).

Class-Mark (Midpoint): The midpoint of each class, obtained by adding the sum of the two limits of the class and dividing by 2.

Formula: X̄ = ΣfX / Σf = ΣfX / n

📌 Example: Using the EPA mileage ratings:

X̄ = 1135.5 / 30 = 37.85

Interpretation: The average mileage rating of the 30 cars is 37.85 miles per gallon.

🔑 Definition — Grouping Error: The error introduced by the assumption that all values falling in a class are equal to the mid-point of the class interval.

The mean from a frequency distribution is an approximation. The true mean from the raw data is 37.82, while the grouped mean is 37.85, showing the error is usually small and never serious for the arithmetic mean.

Desirable Properties of the Arithmetic Mean:

• Best understood average in statistics.
• Relatively easy to calculate.
• Takes into account every value in the series.

Limitation: The mean can be distorted by a few very high or very low values.

📌 Example: Number of floors in buildings: 5, 4, 3, 4, 5, 4, 3, 4, 5, 20, 5, 6, 32, 8, 27. The mean is 9, but 12 out of 15 buildings have 6 floors or less. The skyscrapers have a disproportionate effect on the mean.

The Weighted Arithmetic Mean

The weighted mean is used when data values cannot be regarded as having equal weightage.

Formula: X̄_w = ΣW_i X_i / ΣW_i

📌 Example: Absenteeism in a high school:

X̄_w = 2700 / 300 = 9

The weighted mean correctly shows that 9% of students are absent, unlike the simple average of 8%.

🔑 Definition — Weighted Mean: A mean calculated by assigning a weight to each value, based on its relative importance.

The Median

The median is the middle value of a series when arranged in order of magnitude. It divides the data into two equal halves.

🔑 Definition — Median: A value which divides a set of data into two halves, one half comprising of observations greater than and the other half smaller than it. It is the value at or below which 50% of the data lie.

For Ungrouped Data with an Odd Number of Observations: 📌 Example: Number of floors: 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 6, 8, 20, 27, 32. The median is the 8th value (middle of 15), which is 5. This is more representative than the mean (9).

For Ungrouped Data with an Even Number of Observations: The median is the arithmetic mean of the two middle values. 📌 Example: Passengers on a bus: 4, 9, 14, 18, 23, 47. Median = (14 + 18) / 2 = 16 passengers

For a Discrete Frequency Distribution: To find the median, construct a column of cumulative frequencies. Find the total number of observations (n). The median is the value of the observation at the (n/2)th position. If n is even, the median is the value at the (n/2)th and (n/2 + 1)th position.

📌 Example: Number of pupils per class:

n = 45. The median is the 23rd class. 20 classes have up to 28 pupils, and 28 classes have up to 29 pupils. Thus, the 23rd class has 29 pupils.

⭐ Key Takeaways

1. The mode is the most frequent value, useful for inventory and business decisions but can be non-existent or bi-modal. The arithmetic mean is the sum divided by the number of observations, taking all values into account but susceptible to outliers. The median is the middle value, making it robust to extreme values.
2. For grouped data, the class-mark (midpoint) is used to approximate the arithmetic mean, introducing a small grouping error which is usually negligible.
3. The weighted arithmetic mean is crucial when data points have different levels of importance, assigning weights to each value for a correct overall average.
4. The median is the preferred measure of central tendency for skewed distributions or when outliers are present, as it accurately represents the central location without being distorted.
5. For an even number of raw data points, the median is the average of the two middle values. For a discrete frequency distribution, it is found by locating the middle value using cumulative frequencies.

🧠 Quick Revision Questions

1. What is the formula for the mode in a grouped frequency distribution, and what does each symbol represent?
2. Explain the difference between the arithmetic mean and the weighted arithmetic mean, providing a scenario where the weighted mean is necessary.
3. How is the median calculated for a dataset with an even number of observations, and why is it often preferred over the mean for data with extreme values?
4. What is "grouping error" and why is it generally not a serious issue when calculating the arithmetic mean from grouped data?
5. In a discrete frequency distribution of class sizes, with 45 total classes, how would you determine that the median number of pupils per class is 29?

📘 Lecture 25 — Median, Empirical Relation, Quartiles

📖 Overview: This lecture extends the concept of the median to frequency distributions of continuous variables and open-ended distributions. It introduces the empirical relation between the mean, median, and mode, and then explores the broader family of quantiles—including quartiles, deciles, and percentiles—which allow us to partition data into equal parts. Finally, it demonstrates the graphic location of these measures using an ogive, providing a quick visual method for analysis.

🗂️ Topics Covered

The lecture begins with the formula for the median in a continuous frequency distribution, illustrated with the EPA mileage ratings example. It then addresses the special case of open-ended frequency distributions. The empirical relation between the mean, median, and mode is introduced for moderately skewed, unimodal distributions. The concept of quantiles is formally defined, with formulas for quartiles, deciles, and percentiles. The lecture concludes with the method for the graphic location of these quantiles using a cumulative frequency polygon (ogive).

📝 Lecture Summary

Median in Case of a Frequency Distribution of a Continuous Variable

For a frequency distribution, the median is calculated using a specific interpolation formula. This formula assumes that observations within the median class are evenly distributed between its class boundaries.

🔑 Definition — Median Class: The class interval for which the cumulative frequency is just in excess of n/2.

📐 Formula: $$\tilde{X} = l + \frac{h}{f}\left(\frac{n}{2} - c\right)$$ → This formula gives the median value by starting at the lower boundary of the median class (l) and adding a fraction of the class interval size (h), proportional to how far into the class the median falls. f is the frequency of the median class, n is the total number of observations, and c is the cumulative frequency of the class before the median class.

📌 Example: For the EPA mileage ratings data:

• n = 30, n/2 = 15. The median class is 36.0 – 38.9 (class boundaries 35.95 – 38.95).
• l = 35.95, h = 3, f = 14, c = 6.
• Applying the formula: $\tilde{X} = 35.95 + \frac{3}{14}(15 - 6) = 35.95 + 1.93 = 37.88 \approx 37.9$ miles per gallon.
• This means half of the cars have mileage less than or up to 37.88 mpg. 💡 Why this matters: The median is preferable to the mean when data contains extreme values, as it is not pulled by outliers.

Median in Case of an Open-Ended Frequency Distribution

An open-ended frequency distribution has classes like "Less than 2000" or "5000 and above". One major advantage of the median in such cases is that, unless the median itself falls into an open-ended class, there is no need to estimate the missing boundary.

📌 Example: The distribution of workers' wages:

Empirical Relation Between Mean, Median and Mode

This is an observational relationship that holds for unimodal distributions of moderate skewness. In such distributions, the median is sandwiched between the mean and the mode. The key observation is that the distance between the mode and the median is roughly twice the distance between the median and the mean.

🔑 Definition — Empirical Relation: An approximate relationship based on observation, not a rigid mathematical law.

📐 Formula: Mode = 3 Median – 2 Mean → This is derived from the relationship: Median – Mode = 2 (Mean – Median), which simplifies to the formula above.

💡 Why this matters: In a perfectly symmetrical distribution, the mean, median, and mode are all equal. This empirical relation quantifies how they separate in a skewed distribution. It does not hold for J-shaped or extremely skewed distributions.

Quartiles, Deciles & Percentiles

These are collectively known as quantiles (or fractiles) and are used to partition a frequency distribution into equal parts, allowing for a more detailed analysis of data position. The median is the second quartile, the 5th decile, and the 50th percentile.

🔑 Definition — Quartiles: The three values ((Q_1, Q_2, Q_3)) that divide the total area under a frequency polygon into four equal parts (25% each). 🔑 Definition — Deciles: The nine values ((D_1) to (D_9)) that divide the total area into ten equal parts. 🔑 Definition — Percentiles: The 99 values ((P_1) to (P_{99})) that divide the total area into 100 equal parts.

📐 Formula (General Quantile Formula): $$Q_k = l + \frac{h}{f}\left(\frac{kn}{100} - c\right)$$ → This is a general formula where k is the desired percentile (e.g., k = 25 for (Q_1), k = 50 for median). For deciles, (\frac{kn}{100}) is replaced by (\frac{kn}{10}), and for quartiles, by (\frac{kn}{4}).

📌 Example: Company A’s yearly sales are at the 90th percentile. This means 90% of all companies have less sales than Company A, and only 10% have more sales. This shows how percentiles provide a relative quantitative location for a measurement within a data set. 💡 Why this matters: Percentile ranking is a powerful tool for comparison, but it is only practically valuable for large data sets.

Graphic Location of Quantiles

Quantiles can be located graphically using a cumulative frequency polygon (ogive).

📌 Example (EPA Mileage Ratings):

1. Plot the Ogive: Draw the cumulative frequency polygon for the data.
2. Locate Median:
   • Calculate n/2 = 15.
   • Draw a horizontal line from 15 on the y-axis to the ogive.
   • From the point of intersection, draw a vertical line down to the x-axis.
   • Read the value on the x-axis, which is approximately 37.9, matching the calculated median.
3. Locate Quartiles:
   • For (Q_1): draw a horizontal line from n/4 on the y-axis.
   • For (Q_3): draw a horizontal line from 3n/4 on the y-axis.
4. Locate Deciles & Percentiles: The same process is followed, drawing horizontal lines from the corresponding fraction of n (e.g., n/10, 2n/100).

⭐ Key Takeaways

You must remember the median formula for grouped data and the empirical relation (Mode = 3 Median – 2 Mean) for moderately skewed distributions. The general formula for all quantiles (quartiles, deciles, percentiles) follows the same structure as the median, only changing the fraction of n. The graphic location of any quantile is always done by drawing a horizontal line from the appropriate cumulative value on the y-axis to the ogive, then dropping a vertical line to the x-axis. Remember that median, Q2, D5, and P50 are all the same measure, just different names.

🧠 Quick Revision Questions

1. What are the four components ((l, h, f, c)) of the median formula for a continuous frequency distribution?
2. Write down the empirical relation between the mean, median, and mode for a moderately skewed distribution.
3. What is the difference between a quartile, a decile, and a percentile?
4. Which quantile is equivalent to the 40th percentile and the 4th decile?
5. Describe the process for graphically locating the 3rd quartile ((Q_3)) from a cumulative frequency polygon (ogive).

📘 Lecture 26 — Geometric mean, Harmonic mean & relationship between them

📖 Overview: This lecture introduces two additional measures of central tendency — the geometric mean and harmonic mean — explaining their computation for raw and grouped data. It clarifies when each measure is appropriate, establishes the mathematical relationship between the arithmetic, geometric, and harmonic means, and briefly covers two other measures: mid-range and mid-quartile range.

🗂️ Topics Covered

The lecture covers the geometric mean for raw and grouped data, its use in averaging relative changes, and a detailed example involving firm turnover. It then presents the harmonic mean, its application with a speed-distance-time example, and rules for selecting the appropriate average. Finally, the relationship between the three means is stated, followed by the mid-range and mid-quartile range as other measures of central tendency.

📝 Lecture Summary

GEOMETRIC MEAN:

The geometric mean (G) of a set of n positive values X1, X2,…, Xn is defined as the positive nth root of their product. The formula is G = (X1×X2×…×Xn)^(1/n), where all Xi > 0. For large n, computation becomes cumbersome due to root extraction, so logarithms simplify the process: log G = (∑log X)/n, thus G = antilog( (∑log X)/n ).

🔑 Definition — Geometric Mean: the positive nth root of the product of n positive values. 📐 Formula: log G = (∑log X)/n → G is the antilog of the average of the logs of all X values. 📌 Example: Find GM of the numbers: 45, 32, 37, 46, 39, 36, 41, 48, 36. Step 1: Compute log of each value (base 10): 1.6532, 1.5052, 1.5682, 1.6628, 1.5911, 1.5563, 1.6128, 1.6812, 1.5563. Step 2: Sum of logs = 14.3870. n = 9. Step 3: log G = 14.3870/9 = 1.5986. Step 4: G = antilog(1.5986) = 39.68.

GEOMETRIC MEAN FOR GROUPED DATA:

For a frequency distribution with k classes, midpoints X1,…,Xk, and frequencies f1,…,fk (∑f = n), the geometric mean formula becomes G = (X1^f1 × X2^f2 × … × Xk^fk)^(1/n). In log terms: log G = (∑ f log X)/n, so G = antilog( (∑ f log X)/n ).

📐 Formula: log G = (∑ f log X)/n → the weighted average of logarithms, weighted by frequency. 📌 Example: EPA mileage ratings data (30 cars). Class marks (X): 31.45, 34.45, 37.45, 40.45, 43.45. Their logs: 1.4976, 1.5372, 1.5735, 1.6069, 1.6380. Frequencies: 2, 4, 14, 8, 2. f log X values: 2.9952, 6.1488, 22.0290, 12.8552, 3.2760. ∑ f log X = 47.3042. n = 30. log G = 47.3042/30 = 1.5768. G = antilog(1.5768) = 37.74 mph.

WHEN TO USE THE GEOMETRIC MEAN:

When relative changes in some variable quantity are averaged, the geometric mean is preferred over the arithmetic mean.

📌 Example: Firm’s turnover (year, turnover, percent of previous year): 1958: £2000; 1959: £2500 (125%); 1960: £5000 (200%); 1961: £7500 (150%); 1962: £10500 (140%). Arithmetic mean of percentages = (125+200+150+140)/4 = 153.75%. Using this, the computed 1962 turnover is £11,176 — incorrect. Geometric mean = (125×200×150×140)^(1/4) = 151.37%. Using this, the computed 1962 turnover is exactly £10,500 — correct. Interpretation: On average, each year’s turnover is 51.37% higher than the previous year. 💡 Why this matters: The arithmetic mean overestimates average growth rates because it ignores the multiplicative nature of successive percentage changes. The geometric mean correctly averages ratios.

HARMONIC MEAN:

The harmonic mean (HM) is defined as the reciprocal of the arithmetic mean of the reciprocals of the values. For raw data: HM = n / ∑(1/X). For grouped data (with midpoints X and frequencies f): HM = n / ∑(f/X).

🔑 Definition — Harmonic Mean: reciprocal of the arithmetic mean of the reciprocals of the data values. 📐 Formula (raw data): HM = n / ∑(1/X) → average of reciprocals, then take reciprocal. 📌 Example: Car travels 10 intervals of 10 miles each at varying speeds (30,35,40,40,45,40,50,55,55,30 mph). Find average speed. Arithmetic mean of speeds = 42 mph. But average speed = total distance (100 miles) / total time. Total time = sum of (10/speed) for each interval = 10/30+10/35+10/40+10/40+10/45+10/40+10/50+10/55+10/55+10/30 = 2.4881 hours. True average speed = 100/2.4881 = 40.2 mph. Harmonic mean: HM = n/∑(1/X) = 10/(0.0333+0.0286+…+0.0333) = 10/0.2488 = 40.2 mph. Correct result.

RULES FOR SELECTING THE APPROPRIATE AVERAGE:

1. When values are given as x per y where x is constant and y is variable, the Harmonic Mean is appropriate.
2. When values are given as x per y where y is constant and x is variable, the Arithmetic Mean is appropriate.
3. When relative changes in a variable are to be averaged, the Geometric Mean is appropriate.

📌 Example (Rule 2): Marks out of 20 for 10 students (13,11,9,9,6,5,19,17,12,9). Denominator (20) is constant, numerator varies. Arithmetic mean = 110/10 = 11. This is correct.

RELATION BETWEEN ARITHMETIC, GEOMETRIC AND HARMONIC MEANS:

For any set of positive values: Arithmetic Mean ≥ Geometric Mean ≥ Harmonic Mean. Equality holds only when all values are equal.

SOME OTHER MEASURES OF CENTRAL TENDENCY:

Mid-Range: If x0 is the smallest observation and xm is the largest, then mid-range = (x0 + xm)/2. It represents a value roughly in the middle of the dataset by averaging extremes.

Mid-Quartile Range (Mid-Hinge): If Q1 is the first quartile and Q3 is the third quartile, then mid-quartile range = (Q1 + Q3)/2. It averages the upper and lower quartiles.

🔑 Definition — Mid-Range: arithmetic mean of the smallest and largest observations. 🔑 Definition — Mid-Quartile Range: arithmetic mean of the first and third quartiles; also called mid-hinge.

⭐ Key Takeaways

The geometric mean is essential when averaging ratios or percentage changes because it correctly captures multiplicative growth, while the harmonic mean is necessary when averaging rates where the numerator is fixed (like constant distance with variable speed). Arithmetic mean overestimates average growth rates, so the choice of measure critically affects results. The relationship A.M. ≥ G.M. ≥ H.M. always holds for positive data, and the mid-range and mid-quartile range offer simpler alternatives for a quick center estimate. Students must apply the three rules to select the correct average in any problem.

🧠 Quick Revision Questions

1. Why does computing the geometric mean of raw data use logarithms, and what is the formula?
2. A firm’s annual profits are 120%, 150%, 110%, and 130% of the previous year. Which average correctly computes the average annual growth? Show the calculation.
3. If you have a speed of 60 km/h for the first 50 km and 40 km/h for the next 50 km, what is the correct average speed? Which mean gives it?
4. State the relationship between arithmetic, geometric, and harmonic means for positive unequal values.
5. Define the mid-range and mid-quartile range. When might they be used instead of the arithmetic mean?

📘 Lecture 27 — Measures of Dispersion

📖 Overview: This lecture introduces the concept of dispersion as a measure of variability in data, distinguishing it from measures of central tendency. It covers absolute and relative measures, specifically the range and quartile deviation, along with their coefficients, and explains why measuring spread is critical for data analysis.

🗂️ Topics Covered

The lecture begins with the concept of dispersion and why it is needed alongside averages, using examples of age and class-size distributions to illustrate that identical means can hide very different spreads. It then defines absolute versus relative measures, followed by detailed explanations of the range and its coefficient of dispersion, and the quartile deviation and its coefficient of quartile deviation, including their calculations, graphical representations, advantages, and disadvantages. The lecture concludes by introducing the mean deviation and standard deviation as measures based on all data values.

📝 Lecture Summary

Concept of dispersion

Just as variable series differ in their location (average), they also differ in the amount of variability they exhibit. We need a measure of variability or dispersion to accompany the relevant measure of position or ‘average’, as identical averages can come from entirely different distributions.

🔑 Definition — Dispersion: The amount of variability or spread present in a set of data.

💡 Why this matters: Two datasets can have the same mean but very different distributions, so dispersion measures are essential for a complete description.

Absolute and relative measures of dispersion

There are two types of measurements of dispersion: absolute and relative. An absolute measure of dispersion measures dispersion in the same units as the data. A relative measure of dispersion is expressed as a ratio, coefficient, or percentage and is independent of units.

🔑 Definition — Absolute measure of dispersion: A measure that is expressed in the same units as the original data (e.g., rupees, meters, kilograms).

🔑 Definition — Relative measure of dispersion: A measure that is expressed as a ratio, coefficient, or percentage, making it unitless and useful for comparing datasets of different natures.

Range

The range is defined as the difference between the two extreme values of a dataset, i.e., R = Xm – X0 where Xm is the highest value and X0 the lowest.

📐 Formula: Range = Xm – X0 → The highest value minus the lowest value in the dataset.

Coefficient of dispersion

The coefficient of dispersion is the relative measure of the range, defined as half the range divided by the mid-range.

📐 Formula: Coefficient of Dispersion = (Xm – X0) / (Xm + X0)

This is a pure (dimensionless) number used for comparison purposes. For example, if the coefficient is 0.6 for one dataset and 0.4 for another, there is greater dispersion in the first.

Quartile deviation

The quartile deviation (Q.D.) is defined as half of the difference between the third and first quartiles.

📐 Formula: Q.D. = (Q₃ – Q₁) / 2

It is also known as the semi-interquartile range. The range “Median ± Q.D.” contains approximately 50% of the data.

📌 Example: For Company X: Q₁ = 60 shares, Median = 185 shares, Q₃ = 270 shares. Q.D. = (270 – 60) / 2 = 105 shares. For Company Y: Q₁ = 165 shares, Median = 185 shares, Q₃ = 210 shares. Q.D. = (210 – 165) / 2 = 22 shares. This indicates greater concentration of shareholders about the median in Company Y compared to Company X.

Coefficient of quartile deviation

The coefficient of quartile deviation is the relative measure of the quartile deviation, defined as the quartile deviation divided by the mid-quartile range.

📐 Formula: Coefficient of Quartile Deviation = (Q₃ – Q₁) / (Q₃ + Q₁)

This is a pure number used for comparing variation in two or more sets of data.

⭐ Key Takeaways

The range is the simplest measure of dispersion but is unreliable because it only considers the two extreme values, ignoring all intermediate data. The quartile deviation is superior to the range as it is not affected by extreme values and, together with the median, describes the spread of the middle 50% of data. Both the range and quartile deviation are associated with specific averages: the range with the mid-range and the quartile deviation with the median. Relative measures like the coefficient of dispersion and coefficient of quartile deviation are unitless and essential for comparing variability between different datasets. The mean deviation and standard deviation were introduced as the next measures that use all data values, overcoming the limitation of the range and quartile deviation.

🧠 Quick Revision Questions

1. Why is a measure of dispersion needed even when the mean of two datasets is identical?
2. What is the difference between an absolute and a relative measure of dispersion?
3. Calculate the range for the dataset: 17, 18, 18, 19, 18, 19, 19, 18, 17, 18, 18.
4. If Q₁ = 25 and Q₃ = 75, what is the quartile deviation, and what approximate percentage of data lies between the median minus this value and the median plus this value?
5. How is the coefficient of quartile deviation calculated, and why is it useful?

📘 Lecture 28 — Mean Deviation, Standard Deviation and Variance & Coefficient of variation

📖 Overview: This lecture introduces key measures of dispersion that involve every data value in a dataset. Unlike the range and quartile deviation, the mean deviation and standard deviation consider all observations and are essential for understanding the spread of data around the arithmetic mean, with the standard deviation being the most important measure in statistics.

🗂️ Topics Covered

The lecture covers the mean deviation as a measure of dispersion (for raw data and grouped data), moving to the standard deviation and variance, including their definitions, formulas, and short-cut computation methods. It also addresses relative measures of dispersion including the coefficient of mean deviation, coefficient of standard deviation, and the coefficient of variation.

📝 Lecture Summary

Mean Deviation

The mean deviation is a measure of dispersion based on the absolute deviations of values from their mean. For raw data, the formula is:

M.D. = Σ|dᵢ| / n

For the fatalities dataset (4, 6, 2, 0, 3, 5, 8), the mean is 4. The absolute deviations are 0, 2, 2, 4, 1, 1, 4. Their sum is 14, so M.D. = 14/7 = 2.

🔑 Definition — Mean Deviation: the average of the absolute deviations of observations from their arithmetic mean.

📐 Formula (raw data): M.D. = Σ|xᵢ - x̄| / n → the average absolute distance from the mean

📐 Formula (grouped data): M.D. = Σfᵢ|xᵢ - x̄| / n = Σfᵢ|dᵢ| / n

The mean deviation is also called the mean absolute deviation. While quick and simple, it involves "artificiality" by ignoring algebraic signs of deviations. 💡 Why this matters: Ignoring signs prevents further theoretical development and mathematical applications.

Co-efficient of Mean Deviation

The coefficient of mean deviation is a relative measure obtained by dividing the mean deviation by the average used. When using the mean: Coefficient of M.D. = M.D. / Mean. When using the median: Mean deviation = Σ|xᵢ - x̃| / n and Coefficient of M.D. = M.D. / Median. The median is preferred when the dataset contains very high or very low values.

Standard Deviation and Variance

The standard deviation overcomes the mathematical limitation of the mean deviation by squaring deviations instead of taking absolute values. The variance is the average of the squared deviations:

Variance = Σ(x - x̄)² / n

For the fatalities dataset, Σ(x - x̄)² = 0 + 4 + 4 + 16 + 1 + 1 + 16 = 42, and Variance = 42/7 = 6 (in "squared fatalities").

🔑 Definition — Variance: the average of the squared deviations of observations from their arithmetic mean 🔑 Definition — Standard Deviation: the positive square root of the variance, giving a measure in the original unit of measurement

📐 Formula (raw data): S = √[Σ(x - x̄)² / n]

For the example: S = √6 = 2.45 fatalities.

📐 Formula — Short Cut Formula (raw data): S = √[ (Σx²/n) - (Σx/n)² ]

This uses only Σx and Σx², reducing calculations.

For the example: Σx = 28, Σx² = 154, n = 7. S = √[154/7 - (28/7)²] = √[22 - 16] = √6 = 2.45.

📐 Formula — Standard Deviation (grouped data): S = √[ Σf(x - x̄)² / n ] 📐 Formula — Short Cut Formula (grouped data): S = √[ (Σfx²/n) - (Σfx/n)² ]

📌 Example (electric light bulbs): Life (0-5): f=4, x=2.5; (5-10): f=9, x=7.5; (10-20): f=38, x=15; (20-40): f=33, x=30; (40+): f=16, x=50. Σf=100, Σfx=2437.5, Σfx²=78781.25. S = √[78781.25/100 - (2437.5/100)²] = √[787.8125 - 594.1406] = √193.6719 = 13.9 hundred hours = 1390 hours.

Coefficient of Variation

The coefficient of standard deviation is: S.D. / Mean. Multiplying by 100 gives the coefficient of variation: C.V. = (S / x̄) × 100

🔑 Definition — Coefficient of Variation (C.V.): a relative measure of dispersion expressed as a percentage, useful for comparing variability between different variables or between datasets with very different means.

📌 Example 1 (factory earnings): Country 1: mean = $19.50, S = $4 → C.V. = (4/19.5)×100 = 20.5%. Country 2: mean = Rs. 75, S = Rs. 28 → C.V. = (28/75)×100 = 37.3%. Conclusion: Country 2 has greater variability in earnings.

📌 Example 2 (crop yield): Untreated land: mean = 35 bushels, S = 10 → C.V. = 28.57%. Treated land: mean = 58 bushels, S = 10 → C.V. = 17.24%. Conclusion: The new fertilizer has reduced relative variability.

⭐ Key Takeaways

The mean deviation provides a simple measure of dispersion by averaging absolute deviations from the mean, but its mathematical limitation is ignoring signs. The standard deviation and variance overcome this by squaring deviations, making them the preferred measures of dispersion in statistics. The coefficient of variation is the essential relative measure for comparing variability between different datasets, especially when their means differ substantially or when comparing different variables.

🧠 Quick Revision Questions

1. Why is the sum of deviations from the arithmetic mean always zero, and how does the mean deviation solve this problem?
2. What is the mathematical advantage of the standard deviation over the mean deviation?
3. Compute the standard deviation for the dataset: 2, 4, 6, 8 using the short-cut formula.
4. In which situation would you use the coefficient of variation instead of the standard deviation?
5. If two datasets have the same standard deviation but different means, which dataset has greater relative variability?

📘 Lecture 29 — Permutations and Combinations

📖 Overview: This lecture introduces fundamental counting rules used in probability calculations, specifically permutations and combinations. It then transitions into the foundational concepts of probability theory, including random experiments, sample spaces, and different types of events. Understanding these concepts is crucial for defining and calculating probabilities in various real-world and mathematical scenarios.

🗂️ Topics Covered

This lecture covers the rule of permutation for ordered selections, including factorials and permutations of non-distinct objects. It then explains the rule of combination for unordered selections and its properties as a binomial coefficient. The second half of the lecture introduces random experiments, sample spaces, and various types of events such as simple, compound, complementary, mutually exclusive, exhaustive, and equally likely events.

📝 Lecture Summary

COUNTING RULES:

As discussed in the last lecture, there are certain rules that facilitate the calculations of probabilities in certain situations. They are known as counting rules and include concepts of: 1) Multiple Choice, 2) Permutations, and 3) Combinations. We have already discussed the rule of multiplication in the last lecture.

RULE OF PERMUTATION:

A permutation is any ordered subset from a set of n distinct objects. For example, if we have the set {a, b}, then one permutation is ab, and the other permutation is ba. The number of permutations of r objects, selected in a definite order from n distinct objects is denoted by the symbol nPr, and is given by nPr = n (n – 1) (n – 2) …(n – r + 1) = n! / (n – r)!.

🔑 Definition — Permutation: an ordered subset from a set of n distinct objects. 📐 Formula: nPr = n! / (n – r)! → The number of ways to select and arrange r items in order from a set of n distinct items. 📌 Example: A club consists of four members. How many ways are there of selecting three officers: president, secretary and treasurer? It is evident that the order in which 3 officers are to be chosen, is of significance. Thus there are 4 choices for the first office, 3 choices for the second office, and 2 choices for the third office. Hence the total number of ways in which the three offices can be filled is 4 × 3 × 2 = 24. The same result is obtained by applying the rule of permutations: 4P3 = 4! / (4 - 3)! = 4 × 3 × 2 = 24.

FACTORIALS

7! = 7 × 6 × 5 × 4 × 3 × 2 × 1. 6! = 6 × 5 × 4 × 3 × 2 × 1. 1! = 1. Also, we define 0! = 1.

The factorials and permutations are used to describe possible arrangements of distinct objects. A tree diagram can provide an organized way of listing possible arrangements for a permutation problem.

PERMUTATIONS:

In the formula of nPr, if we put r = n, we obtain: nPn = n(n – 1) (n – 2) … 3 × 2 × 1 = n!. I.e. the total number of permutations of n distinct objects, taking all n at a time, is equal to n!.

🔑 Definition — The total number of permutations of n distinct objects, taking all n at a time, is equal to n!. 📌 Example: Suppose that there are three persons A, B & D, and that they wish to have a photograph taken. The total number of ways in which they can be seated on three chairs (placed side by side) is 3P3 = 3! = 6. These are: ABD, ADB, BAD, BDA, DAB, & DBA.

The above discussion pertained to the case when all the objects under consideration are distinct objects. If some of the objects are not distinct, the formula of permutations modifies as given below: The number of permutations of n objects, selected all at a time, when n objects consist of n1 of one kind, n2 of a second kind, …, nk of a kth kind, is P = n! / (n1! n2! ….. nk!), where Σ ni = n.

📐 Formula: Permutations of non-distinct objects: P = n! / (n1! n2! ….. nk!), where Σ ni = n. 📌 Example: How many different (meaningless) words can be formed from the word ‘committee’? In this example: n = 9 (because the total number of letters in this word is 9), n1 = 1 (c), n2 = 1 (o), n3 = 2 (m’s), n4 = 1 (i), n5 = 2 (t’s), n6 = 2 (e’s). Hence, the total number of (meaningless) words (permutations) is: P = 9! / (1! 1! 2! 1! 2! 2!) = (9 × 8 × 7 × 6 × 5 × 4 × 3 × 2 × 1) / (1 × 1 × 2 × 1 × 1 × 2 × 1 × 2 × 1) = 45360.

RULE OF COMBINATION:

A combination is any subset of r objects, selected without regard to their order, from a set of n distinct objects. The total number of such combinations is denoted by the symbol nCr or (n choose r), and is given by (n choose r) = n! / [r! (n – r)!], where r < n.

It should be noted that nPr = r! * (n choose r). In other words, every combination of r objects (out of n objects) generates r! permutations.

🔑 Definition — Combination: any subset of r objects, selected without regard to their order, from a set of n distinct objects. 📐 Formula: (n choose r) = n! / [r! (n – r)!], where r < n. 📌 Example: Suppose we have a group of three persons, A, B, & C. If we wish to select a group of two persons out of these three, the three possible groups are {A, B}, {A, C} and {B, C}. In other words, the total number of combinations of size two out of this set of size three is 3. Now, suppose that our interest lies in forming a committee of two persons, one of whom is to be the president and the other the secretary of a club. The six possible committees are: (A, B), (B, A), (A, C), (C, A), (B, C) & (C, B). In other words, the total number of permutations of two persons out of three is 6. And the point to note is that each of three combinations mentioned earlier generates 2 = 2! permutations.

The quantity (n choose r) or nCr is also called a binomial co-efficient because of its appearance in the binomial expansion of (a + b)^n = Σ (n choose r) a^(n-r) b^r. The binomial co-efficient has two important properties: i) (n choose r) = (n choose n-r), and ii) (n choose n-r) + (n choose r) = (n+1 choose r). Also, it should be noted that (n choose 0) = 1 = (n choose n) and (n choose 1) = n = (n choose n-1).

📌 Example: A three-person committee is to be formed out of a group of ten persons. In how many ways can this be done? Since the order in which the three persons of the committee are chosen, is unimportant, it is therefore an example of a problem involving combinations. Thus the desired number of combinations is (10 choose 3) = 10! / (3! 7!) = (10 × 9 × 8 × 7 × 6 × 5 × 4 × 3 × 2 × 1) / (3 × 2 × 1 × 7 × 6 × 5 × 4 × 3 × 2 × 1) = 120. In other words, there are one hundred and twenty different ways of forming a three-person committee out of a group of only ten persons!

📌 Example: In how many ways can a person draw a hand of 5 cards from a well-shuffled ordinary deck of 52 cards? The total number of ways of doing so is given by (52 choose 5) = (52 × 51 × 50 × 49 × 48) / (5 × 4 × 3 × 2 × 1) = 2,598,960.

Having reviewed the counting rules that facilitate calculations of probabilities in a number of problems, let us now begin the discussion of concepts that lead to the formal definitions of probability. The first concept in this regard is the concept of Random Experiment. The term experiment means a planned activity or process whose results yield a set of data. A single performance of an experiment is called a trial. The result obtained from an experiment or a trial is called an outcome.

RANDOM EXPERIMENT:

An experiment which produces different results even though it is repeated a large number of times under essentially similar conditions is called a Random Experiment. The tossing of a fair coin, the throwing of a balanced die, drawing of a card from a well-shuffled deck of 52 playing cards, selecting a sample, etc. are examples of random experiments.

A random experiment has three properties: i) The experiment can be repeated, practically or theoretically, any number of times. ii) The experiment always has two or more possible outcomes. An experiment that has only one possible outcome, is not a random experiment. iii) The outcome of each repetition is unpredictable, i.e. it has some degree of uncertainty. 💡 Why this matters: These three properties help identify whether a process qualifies as a random experiment, which is the basis for studying probability. For example, interviewing a person to find out whether or not he or she is a smoker is an example of a random experiment because it fulfills all three properties.

SAMPLE SPACE:

A set consisting of all possible outcomes that can result from a random experiment (real or conceptual), can be defined as the sample space for the experiment and is denoted by the letter S. Each possible outcome is a member of the sample space, and is called a sample point in that space.

📌 Example-1: The experiment of tossing a coin results in either of the two possible outcomes: a head (H) or a tail (T). The sample space for this experiment may be expressed in set notation as S = {H, T}. ‘H’ and ‘T’ are the two sample points. 📌 Example-2: The sample space for tossing two coins once (or tossing a coin twice) will contain four possible outcomes denoted by S = {HH, HT, TH, TT}. 📌 Example-3: The sample space S for the random experiment of throwing two six-sided dice can be described by the Cartesian product A × A, where A = {1, 2, 3, 4, 5, 6}. In other words, S = A × A = {(x, y) | x ∈ A and y ∈ A}, where x denotes the number of dots on the upper face of the first die, and y denotes the number of dots on the upper face of the second die. Hence, S contains 36 outcomes or sample points.

EVENTS:

Any subset of a sample space S of a random experiment, is called an event. In other words, an event is an individual outcome or any number of outcomes (sample points) of a random experiment.

SIMPLE & COMPOUND EVENTS: An event that contains exactly one sample point, is defined as a simple event. A compound event contains more than one sample point, and is produced by the union of simple events. 📌 Example: The occurrence of a 6 when a die is thrown, is a simple event, while the occurrence of a sum of 10 with a pair of dice, is a compound event, as it can be decomposed into three simple events (4, 6), (5, 5) and (6, 4).

OCCURRENCE OF AN EVENT: An event A is said to occur if and only if the outcome of the experiment corresponds to some element of A.

COMPLEMENTARY EVENT: The event “not-A” is denoted by Ā or A^c and called the negation (or complementary event) of A. 📌 Example: If we toss a coin once, then the complement of “heads” is “tails”. If we toss a coin four times, then the complement of “at least one head” is “no heads”.

A sample space consisting of n sample points can produce 2^n different subsets (or simple and compound events). 📌 Example: Consider a sample space S containing 3 sample points, i.e. S = {a, b, c}. Then the 2^3 = 8 possible subsets are φ, {a}, {b}, {c}, {a, b}, {a, c}, {b, c}, {a, b, c}. Each of these subsets is an event. The subset {a, b, c} is the sample space itself and is also an event. It always occurs and is known as the certain or sure event. The empty set φ is also an event, sometimes known as impossible event, because it can never occur.

MUTUALLY EXCLUSIVE EVENTS:

Two events A and B of a single experiment are said to be mutually exclusive or disjoint if and only if they cannot both occur at the same time i.e. they have no points in common. 📌 Example-1: When we toss a coin, we get either a head or a tail, but not both at the same time. The two events head and tail are therefore mutually exclusive. 📌 Example-2: When a die is rolled, the events ‘even number’ and ‘odd number’ are mutually exclusive as we can get either an even number or an odd number in one throw, not both at the same time. Three or more events originating from the same experiment are mutually exclusive if pair wise they are mutually exclusive. If the two events can occur at the same time, they are not mutually exclusive, e.g., if we draw a card from an ordinary deck of 52 playing cars, it can be both a king and a diamond. Therefore, kings and diamonds are not mutually exclusive.

EXHAUSTIVE EVENTS:

Events are said to be collectively exhaustive, when the union of mutually exclusive events is equal to the entire sample space S. 📌 Example: In the coin-tossing experiment, ‘head’ and ‘tail’ are collectively exhaustive events. In the die-tossing experiment, ‘even number’ and ‘odd number’ are collectively exhaustive events.

PARTITION OF THE SAMPLE SPACE: A group of mutually exclusive and exhaustive events belonging to a sample space is called a partition of the sample space. With reference to any sample space S, events A and Ā form a partition as they are mutually exclusive and their union is the entire sample space.

EQUALLY LIKELY EVENTS:

Two events A and B are said to be equally likely, when one event is as likely to occur as the other. In other words, each event should occur in equal number in repeated trials. 📌 Example: When a fair coin is tossed, the head is as likely to appear as the tail, and the proportion of times each side is expected to appear is 1/2. If a card is drawn out of a deck of well-shuffled cards, each card is equally likely to be drawn, and the probability that any card will be drawn is 1/52.

⭐ Key Takeaways

Students must remember the distinction between permutations (ordered arrangements) and combinations (unordered selections), as well as their respective formulas: nPr = n!/(n-r)! and nCr = n!/[r!(n-r)!]. A random experiment has three essential properties: repeatability, multiple possible outcomes, and unpredictable results. The sample space contains all possible outcomes, and any subset of it is an event. Mutually exclusive events cannot occur simultaneously, exhaustive events cover the entire sample space, and equally likely events have the same chance of occurring.

🧠 Quick Revision Questions

1. A club has 10 members. How many different ways can they select a president, vice president, and secretary (order matters)?
2. From a group of 8 people, how many different committees of 3 can be formed?
3. What are the three properties that define a random experiment?
4. When rolling a single six-sided die, what is the sample space? Give an example of a simple event and a compound event from this space.
5. Are the events “drawing a heart” and “drawing a king” from a standard deck of cards mutually exclusive? Why or why not?

📘 Lecture 30 — Definitions of Probability

📖 Overview: This lecture introduces the fundamental concept of probability, beginning with essential preliminary concepts such as mutually exclusive, exhaustive, and equally likely events. It then presents the two main approaches to probability: the subjective approach and the objective approach, with detailed coverage of the classical (a priori) definition and its limitations, leading to the relative frequency definition of probability.

🗂️ Topics Covered

The lecture covers mutually exclusive events with examples from coin tossing and dice rolling, exhaustive events and the partition of the sample space, equally likely events, the subjective or personalistic approach to probability, the classical definition of probability with several worked examples including card drawing and coin tossing, the shortcomings of the classical definition, and finally the relative frequency definition of probability.

📝 Lecture Summary

Mutually Exclusive Events

Mutually exclusive or disjoint events are two events A and B of a single experiment that cannot both occur at the same time, meaning they have no points in common. When we toss a coin, we get either a head or a tail, but not both at the same time, so the two events head and tail are mutually exclusive. When a die is rolled, the events 'even number' and 'odd number' are mutually exclusive as we can get either an even number or an odd number in one throw, not both at the same time. Three or more events are mutually exclusive if pairwise they are mutually exclusive. If the two events can occur at the same time, they are not mutually exclusive — for example, if we draw a card from an ordinary deck of 52 playing cards, it can be both a king and a diamond, so kings and diamonds are not mutually exclusive.

🔑 Definition — Mutually Exclusive Events: Two events A and B are mutually exclusive if they cannot both occur at the same time, i.e., A ∩ B = ∅.

Exhaustive Events

Events are said to be collectively exhaustive when the union of mutually exclusive events is equal to the entire sample space S. In the coin-tossing experiment, 'head' and 'tail' are collectively exhaustive events. In the die-tossing experiment, 'even number' and 'odd number' are collectively exhaustive events.

🔑 Definition — Exhaustive Events: A set of events is collectively exhaustive if their union equals the entire sample space S.

Partition of the Sample Space

A group of mutually exclusive and exhaustive events belonging to a sample space is called a partition of the sample space. With reference to any sample space S, events A and Ā form a partition as they are mutually exclusive and their union is the entire sample space.

Equally Likely Events

Two events A and B are said to be equally likely when one event is as likely to occur as the other. In other words, each event should occur in equal number in repeated trials. When a fair coin is tossed, the head is as likely to appear as the tail, and the proportion of times each side is expected to appear is 1/2. If a card is drawn out of a deck of well-shuffled cards, each card is equally likely to be drawn, and the proportion of times each card can be expected to be drawn in a very large number of draws is 1/52.

🔑 Definition — Equally Likely Events: Two events A and B are equally likely when each event occurs in equal number in repeated trials.

Subjective or Personalistic Probability

The subjective or personalistic probability is a measure of the strength of a person's belief regarding the occurrence of an event A. Probability in this sense is purely subjective and is based on whatever evidence is available to the individual. It has a disadvantage that two or more persons faced with the same evidence may arrive at different probabilities. For example, suppose that a panel of three judges is hearing a trial. It is possible that, based on the evidence that is presented, two of them arrive at the conclusion that the accused is guilty while one of them decides that the evidence is NOT strong enough to draw this conclusion.

🔑 Definition — Subjective Probability: A measure of the strength of a person's belief regarding the occurrence of an event, based on whatever evidence is available to the individual.

Classical Definition of Probability

The Classical or 'A Priori' Definition of Probability states: If a random experiment can produce n mutually exclusive and equally likely outcomes, and if m out of these outcomes are considered favorable to the occurrence of a certain event A, then the probability of the event A, denoted by P(A), is defined as the ratio m/n.

📐 Formula: P(A) = m/n = (Number of favourable outcomes)/(Total number of possible outcomes)

Example 1: If a card is drawn from an ordinary deck of 52 playing cards, find the probability that (i) the card is a red card, (ii) the card is a 10.

Solution: The total number of possible outcomes is 52, and we assume all possible outcomes are equally likely.

(i) Let A represent the event that the card drawn is a red card. The number of outcomes favourable to event A is 26 (13 diamonds + 13 hearts). Hence P(A) = 26/52 = 1/2.

(ii) Let B represent the event that the card drawn is a 10. There are 4 tens (one in each suit). Hence P(B) = 4/52 = 1/13.

Example 2: A fair coin is tossed three times. What is the probability that at least one head appears?

Solution: The sample space is S = {HHH, HHT, HTH, THH, HTT, THT, TTH, TTT}, so n(S) = 8. Let A denote the event that at least one head appears. Then A = {HHH, HHT, HTH, THH, HTT, THT, TTH} and n(A) = 7. Hence P(A) = n(A)/n(S) = 7/8.

Example 3: Four items are taken at random from a box of 12 items and inspected. The box is rejected if more than 1 item is found to be faulty. If there are 3 faulty items in the box, find the probability that the box is accepted.

Solution: The sample space S contains C(12,4) = 495 sample points. The box contains 3 faulty and 9 good items. The box is accepted if there is (i) no faulty items, or (ii) one faulty item in the sample of 4 items selected.

Let A denote the event the number of faulty items chosen is 0 or 1. Then: n(A) = C(3,0) × C(9,4) + C(3,1) × C(9,3) = 126 + 252 = 378 sample points.

Therefore, P(A) = 378/495 = 0.76.

Hence, the probability that the box is accepted is 76% (in spite of the fact that the box contains 3 faulty items).

Shortcomings of the Classical Definition

The classical definition has the following shortcomings: i) This definition is said to involve circular reasoning as the term "equally likely" really means "equally probable". Thus probability is defined by introducing concepts that presume a prior knowledge of the meaning of probability. ii) This definition becomes vague when the possible outcomes are infinite in number, or uncountable. iii) This definition is NOT applicable when the assumption of equally likely does not hold. And the fact of the matter is that there are numerous situations where the assumption of equally likely cannot hold.

Relative Frequency Definition of Probability

The relative frequency definition of probability states: If an experiment is repeated a large number of times under (more or less) identical conditions, and if the event of our interest occurs a certain number of times, then the proportion in which this event occurs is regarded as the probability of that event.

For example, a large number of students sit for the matric examination every year. Since the total number of students appearing is very large, the proportion of students who obtain the first division can be regarded as the probability of obtaining the first division, and the proportion of students who obtain the second division can be regarded as the probability of obtaining the second division, and so on.

💡 Why this matters: The relative frequency definition overcomes the limitations of the classical definition by not requiring the assumption of equally likely outcomes, making it applicable to real-world situations where outcomes are not equally probable.

⭐ Key Takeaways

The three essential preliminary concepts for understanding probability are mutually exclusive events (cannot occur together), exhaustive events (union equals sample space), and equally likely events (same chance of occurrence). The subjective approach to probability depends on personal belief and can vary between individuals. The classical definition P(A) = m/n requires mutually exclusive and equally likely outcomes, but suffers from circular reasoning and inapplicability when assumptions fail. The relative frequency definition overcomes these limitations by basing probability on long-run proportions from repeated experiments.

🧠 Quick Revision Questions

1. What are mutually exclusive events? Give an example involving a deck of cards.
2. What is meant by a partition of the sample space?
3. State the classical definition of probability and write its formula.
4. What are the three main shortcomings of the classical definition of probability?
5. How does the relative frequency definition of probability overcome the limitations of the classical definition?

📘 Lecture 31 — Relative Frequency, Axiomatic Definition & Laws of Probability

📖 Overview: This lecture introduces the relative frequency (empirical) definition of probability, which is based on the long-run behavior of events observed in repeated experiments. It then presents the formal axiomatic definition of probability, followed by key laws including the rule of complementation and the addition theorem. These concepts are essential for understanding how probability is quantified and applied in real-world situations where outcomes are not equally likely.

🗂️ Topics Covered

This lecture covers three main definitions of probability: the relative frequency ('a posteriori') definition based on the limit of relative frequency as the number of trials increases, the axiomatic definition based on Kolmogorov's three axioms, and the classical ('a priori') definition. It also discusses non-quantifiable (inductive) probability and presents the rule of complementation and the addition theorem for non-mutually exclusive events, illustrated with examples from coin-tossing and birth statistics.

📝 Lecture Summary

The Relative Frequency Definition of Probability (‘A Posteriori’ Definition of Probability)

If a random experiment is repeated a large number of times, say n times, under identical conditions, and if an event A is observed to occur m times, then the probability of event A is defined as the limit of the relative frequency m/n as n tends to infinity. Symbolically, P(A) = Lim (m/n) as n→∞. This definition assumes that as n increases indefinitely, the ratio m/n stabilizes at the numerical value P(A).

This definition is also called empirical probability because it is based on empirical evidence (observational data), or statistical probability, as it forms the basis of mathematical statistics. It is very useful in practical situations where the classical definition cannot be applied because outcomes are NOT equally likely.

🔑 Definition — Relative Frequency Definition of Probability: The probability of an event A is the limit of the relative frequency (m/n) as the number of trials (n) approaches infinity. 📐 Formula: P(A) = Lim (m/n) as n→∞ → The probability equals the long-run proportion of times the event occurs.

📌 Example-1 — Coin-Tossing (Kerrich, 1946): A coin was tossed 10,000 times, yielding 5067 heads and 4933 tails. The proportion of heads fluctuated widely at first but settled down near 0.5 as the number of tosses increased. The hypothetical limiting value (near 0.5) is the statistical probability of heads.

📌 Example-2 — Sex Ratio at Birth: Laplace recorded that among 215,599 births in France (1800-1802), there were 110,312 boys and 105,287 girls, giving proportions of 0.512 and 0.488. For major regions of England (each based on ~100,000 births), the proportion of male births ranged between 0.512 and 0.517 only. However, for rural districts of Dorset (each based on ~200 births), the proportion fluctuated between 0.38 and 0.59. The larger sample size leads to greater constancy. The hypothetical limiting value (near 0.514) is the statistical probability of a male birth.

💡 Why this matters: The relative frequency definition applies when we need to compute probability numerically but the classical definition cannot be used because outcomes are not equally likely.

Non-Quantifiable (Inductive) Probability

This refers to the degree of belief which it is reasonable to place in a proposition on given evidence. It is difficult to express inductive probabilities numerically. Most statisticians conclude that inductive probability cannot be measured and therefore cannot be used in the mathematical theory of statistics. For this reason, usual methods of statistical inference (tests of significance, confidence intervals) are based entirely on the concept of statistical probability.

The Axiomatic Definition of Probability

Introduced in 1933 by the Russian mathematician Andrei N. Kolmogorov, this definition is based on a set of axioms. Let S be a sample space with sample points E1, E2, ... En. To each sample point, we assign a real number P(Ei), called the probability of Ei, that must satisfy the following basic axioms:

Axiom 1: For any event Ei, 0 < P(Ei) < 1. Axiom 2: P(S) = 1 for the sure event S. Axiom 3: If A and B are mutually exclusive events (subsets of S), then P(A ∪ B) = P(A) + P(B).

According to the axiomatic theory, some probability defined as a non-negative real number is attached to each sample point Ei such that the sum of all such numbers must equal ONE. The assignment of probabilities may be based on past evidence (empirical probability) or on underlying conditions ensuring equally likely outcomes (classical probability).

🔑 Definition — Axiomatic Definition of Probability: A set of axioms (Kolmogorov's axioms) that any valid probability assignment must satisfy: non-negativity, the probability of the sample space equals 1, and additivity for mutually exclusive events.

📌 Example — Births in England and Wales (1956): From Table-1 data (716,740 births), relative frequencies are computed as probabilities in Table-2. For example, the probability of a male birth, P(M) = P(Male livebirth or Male stillbirth) = P(A) + P(B) = .5021 + .0120 = .5141. Similarly, P(Stillbirth) = P(B or D) = .0120 + .0109 = .0229.

Law of Complementation

If Ā is the complement of an event A relative to the sample space S, then P(Ā) = 1 - P(A). Hence, the probability of the complement of an event is equal to one minus the probability of the event. Complementary probabilities are very useful for questions like "What is the probability that, in tossing two fair dice, at least one even number will appear?"

🔑 Definition — Rule of Complementation: The probability that an event does NOT occur equals 1 minus the probability that it does occur. 📐 Formula: P(Ā) = 1 - P(A)

📌 Example — Coin tossed 4 times: What is the probability that at least one head occurs? The sample space has 2⁴ = 16 equally likely outcomes. Instead of finding P(at least one head) directly, find P(no head) = P(Ā) where Ā = {TTTT}. P(Ā) = 1/16. Then P(A) = 1 - P(Ā) = 1 - 1/16 = 15/16.

Addition Law (General Addition Theorem of Probability)

If A and B are any two events defined in a sample space S, then P(A ∪ B) = P(A) + P(B) - P(A ∩ B). In words: "If two events A and B are not mutually exclusive, then the probability that at least one of them occurs is given by the sum of the separate probabilities of events A and B minus the probability of the joint event A ∩ B."

🔑 Definition — Addition Theorem: For any two events A and B, the probability of A or B occurring equals the sum of their individual probabilities minus the probability of both occurring. 📐 Formula: P(A ∪ B) = P(A) + P(B) - P(A ∩ B)

⭐ Key Takeaways

The relative frequency definition of probability defines probability as the limiting value of the relative frequency as the number of trials approaches infinity, making it applicable to real-world situations with unequally likely outcomes. The axiomatic definition, based on Kolmogorov's three axioms, provides the formal mathematical foundation for probability theory. The rule of complementation (P(Ā) = 1 - P(A)) simplifies probability calculations, especially for "at least one" problems. The addition theorem (P(A ∪ B) = P(A) + P(B) - P(A ∩ B)) accounts for overlapping events. Non-quantifiable inductive probability, based on degree of belief, is not used in mathematical statistics, which relies on statistical probability instead.

🧠 Quick Revision Questions

1. What is the relative frequency definition of probability, and what condition must be satisfied as n increases?
2. State Kolmogorov's three axioms of probability.
3. Using the law of complementation, if P(no head in 4 coin tosses) = 1/16, what is P(at least one head)?
4. Write the addition theorem formula for two events A and B that are not mutually exclusive.
5. Why is inductive (non-quantifiable) probability generally not used in mathematical statistics?

📘 Lecture 32 — Independent and Dependent Events, Multiplication Theorem & Marginal Probability

📖 Overview: This lecture explores the distinction between independent and dependent events in probability, introduces the multiplication theorem for independent events, and explains the concept of marginal probability. It provides practical examples to demonstrate how to determine event independence and calculate joint, conditional, and marginal probabilities, which are fundamental for statistical analysis.

🗂️ Topics Covered

The lecture begins by reviewing the addition and multiplication theorems through a comprehensive example involving ball transfers between bags. It then defines independent and dependent events, presents the special case of the multiplication theorem for independent events, and proves this theorem. Using dice and birth statistics examples, the lecture demonstrates how to verify independence and introduces marginal probability by examining how joint probabilities sum to marginal values in contingency tables.

📝 Lecture Summary

Review of Addition and Multiplication Theorems

The lecture opens with a comprehensive example involving two bags of colored balls. Bag A contains 10 white and 3 black balls, while Bag B contains 3 white and 5 black balls. Two balls are transferred from Bag A to Bag B, then one ball is drawn from Bag B. This problem requires both the addition and multiplication theorems to find the probability that the final ball is white.

The three mutually exclusive ways for the transfer event A to occur are: A1 (2 white balls transferred), A2 (1 white and 1 black ball transferred), and A3 (2 black balls transferred). Their probabilities are calculated as:

• P(A1) = C(10,2) ÷ C(13,2) = 45/78
• P(A2) = C(10,1)×C(3,1) ÷ C(13,2) = 30/78
• P(A3) = C(3,2) ÷ C(13,2) = 3/78

After each transfer scenario, the composition of Bag B changes, affecting the conditional probability of drawing white:

• After A1: Bag B has 5 white, 5 black → P(W/A1) = 5/10
• After A2: Bag B has 4 white, 6 black → P(W/A2) = 4/10
• After A3: Bag B has 3 white, 7 black → P(W/A3) = 3/10

Using the multiplication theorem: P(Ai ∩ W) = P(Ai) × P(W/Ai)

• P(A1 ∩ W) = 45/78 × 5/10 = 15/52
• P(A2 ∩ W) = 30/78 × 4/10 = 2/13
• P(A3 ∩ W) = 3/78 × 3/10 = 3/260

Finally, the addition theorem gives: P(W) = 15/52 + 2/13 + 3/260 = 59/130 = 0.45

Independent and Dependent Events

Two events A and B in the same sample space S are defined to be independent (or statistically independent) if the probability that one event occurs is not affected by whether the other event has or has not occurred.

🔑 Definition — Independent Events: Events A and B are independent if and only if P(A/B) = P(A) and P(B/A) = P(B).

📐 Formula — Special Multiplication Theorem for Independent Events: P(A ∩ B) = P(A) × P(B) Plain-English meaning: For independent events, the probability that both occur is simply the product of their individual probabilities.

Rationale: The general multiplication theorem states P(A ∩ B) = P(A) × P(B/A). Substituting P(B/A) = P(B) yields the special case for independent events.

Two events A and B are defined to be dependent if P(A ∩ B) ≠ P(A) × P(B). This means the occurrence of one event in some way affects the probability of the other event occurring.

💡 Why this matters: Two events that are independent can never be mutually exclusive. Mutually exclusive events cannot occur simultaneously (P(A∩B)=0), while independent events require P(A∩B)=P(A)P(B), which can only be zero if at least one event has zero probability.

📌 Example — Dice Independence: Two fair dice, one red and one green, are thrown. Let A = red die shows an even number, and B = green die shows a 5 or 6.

• Sample space S has 36 outcomes
• P(A) = 3/6, P(B) = 2/6
• P(A ∩ B) = 6/36 (outcomes: (2,5), (4,5), (6,5), (2,6), (4,6), (6,6))
• P(A) × P(B) = 3/6 × 2/6 = 6/36 = P(A ∩ B)
• Therefore, events A and B are independent.

📌 Example — Birth Statistics Dependency: Using data from England and Wales (1956) on births by sex and whether liveborn or stillborn:

• P(stillbirth/male birth) = 8,609/368,490 = 0.0234
• P(stillbirth/female birth) = 7,796/348,258 = 0.0224
• Overall P(stillbirth) = 16,405/716,740 = 0.0229

Since the conditional probabilities differ from the unconditional probability, sex and stillbirth are statistically dependent. The sex of a baby has a small effect on its chance of being stillborn.

Marginal Probability

Marginal probabilities are the probabilities that appear in the margins of a contingency table.

Using the birth statistics data (Table-2):

🔑 Definition — Marginal Probability: The probability of a single event (e.g., male birth, stillbirth) obtained by summing joint probabilities across the categories of another variable.

The marginal probability of male birth (0.5141) equals the sum of joint probabilities: P(male and liveborn) + P(male and stillborn) = 0.5021 + 0.0120 = 0.5141. This follows the addition theorem for mutually exclusive events.

The relationship between conditional, joint, and marginal probabilities is:

P(stillbirth/male birth) = P(male birth and stillbirth) ÷ P(male birth)
                         = 0.0120 ÷ 0.5141
                         = 0.0233

⭐ Key Takeaways

The critical distinction between independent and dependent events lies in whether the occurrence of one event affects the probability of another. For independent events, the special multiplication theorem P(A∩B)=P(A)P(B) applies directly, while dependent events require conditional probabilities in the general multiplication theorem. Marginal probabilities, found in table margins, are obtained by summing joint probabilities across categories and represent unconditional probabilities of single events. The relationship P(A/B)=P(A∩B)/P(A) connects conditional, joint, and marginal probabilities. When analyzing real data, independence is rarely perfect, and slight deviations from independence (as shown in the birth statistics example) indicate relationships between variables.

🧠 Quick Revision Questions

1. What condition must be satisfied for two events A and B to be considered statistically independent?
2. Why can independent events never be mutually exclusive?
3. In the dice example, if P(A)=3/6, P(B)=2/6, and P(A∩B)=6/36, how do you verify these events are independent?
4. How is a marginal probability computed from joint probabilities in a contingency table?
5. Using the birth statistics data, explain why sex and stillbirth are considered dependent events despite the small difference in conditional probabilities.