MODULE IV: DESCRIPTIVE STATISTICS

Curricular Scope & Foundational Overview: Module IV presents an exhaustive, graduate-level academic exploration of univariate summary metrics in Fundamentals of Business Mathematics and Statistics (COM2MN104). When raw commercial data is aggregated, decision-makers require quantitative measures that capture its central location, dispersion, and distributional shape. This module covers Measures of Central Tendency (Arithmetic Mean, Weighted Mean, Combined Mean, Geometric Mean, Harmonic Mean, Median, Mode, Quartiles, Deciles, and Percentiles), Measures of Dispersion (Range, Quartile Deviation, Mean Deviation, Standard Deviation, Variance, and the Coefficient of Variation), and the analysis of distributional geometry through Skewness (Karl Pearson's and Bowley's measures) and Kurtosis (Peakedness of curves and moment-based ratios).

Unit 12: Measures of Central Tendency (Statistical Averages)

12.1 The Concept, Purpose, and Requisites of an Ideal Average

A measure of central tendency (or statistical average) is a single representative value that condenses a massive body of quantitative data into a central reference point around which observations cluster. In financial analysis, supply chain management, and executive remuneration, averages serve as the baseline metric for setting corporate benchmarks, projecting quarterly budgets, and comparing organizational units.

According to the classical criteria formulated by statistical pioneer George Udny Yule, an ideal average must satisfy six strict mathematical requirements:

1. Rigidly Defined

The average must be formulated by an unambiguous algebraic definition leaving no room for personal interpretation or investigator subjectivity.

2. Based on All Observations

Every observation in the dataset must contribute to its value. If any single data point changes, the average must change.

3. Readily Comprehensible

It should possess intuitive, common-sense interpretability for business executives and non-technical stakeholders.

4. Simple to Compute

The mathematical calculation should be straightforward, algorithmic, and computationally efficient across large series.

5. Amenable to Algebraic Treatment

The metric must permit further algebraic manipulations (such as computing a combined average or standard deviation across multiple sub-groups).

6. Sampling Stability

It should exhibit minimal fluctuation from one random sample to another drawn from the exact same universe.

12.2 The Arithmetic Mean (AM)

The Arithmetic Mean ( $X ˉ$ ) is the most universally deployed average. It is computed by dividing the sum of all observations by the total number of observations. In grouped continuous series, class intervals are represented by their class mid-points ((m)).

Arithmetic Mean Computational Methods CENTRAL METRICS

D i r ec tM e t h o d : X_{b} a r = (S U M f * m) / N ∣ S t e p - D e v ia t i o n : X_{b} a r = A + [(S U M f * d^{'}) / N] * c

Where:

m = Class midpoint = (Lower boundary + Upper boundary) / 2
f = Class frequency; N = Total frequency ((sum f))
A = Assumed Mean (an arbitrary round midpoint selected near the center)
d' = Step deviation = ((m - A) / c), where c is the common class width

Special Mathematical Properties of the Arithmetic Mean:

Property 1 (Zero Algebraic Sum of Deviations): The algebraic sum of deviations of individual items taken from their arithmetic mean is identically zero:

$\sum (X - X ˉ) = 0] < / l i >< l i >< b > P r o p er t y 2 (L e a s tS u m o f S q u a r e d D e v ia t i o n s) :< / b > T h e \sum o f s q u a r e dd e v ia t i o n s i s ama t h e ma t i c a l minim u m w h e n t ak e n f r o m t h e a r i t hm e t i c m e an r a t h er t hanan y o t h er a r bi t r a r y n u mb er : [\sum (X - X ˉ)^{2} < \sum (X - A)^{2} (for any A e q X ˉ)$
Property 3 (Combined Arithmetic Mean): If two groups have sample sizes (N_1) and (N_2) with means $X ˉ_{1}$ and $X ˉ_{2}$ , the combined mean ( $X ˉ_{12}$ ) of the pooled dataset is:
$X ˉ_{12} = \frac{N _{1} X ˉ _{1} + N _{2} X ˉ _{2}}{N _{1} + N _{2}}$
Property 4 (Weighted Arithmetic Mean): When different items carry unequal commercial importance (weights (w_i)):

$X ˉ_{w} = \frac{\sum w X}{\sum w}$

Numerical Problem 12.1: Calculation of Continuous Mean and Corrected Mean
Part A (Continuous Series): Calculate the mean daily sales (₹ in Thousands) of 50 retail kiosks using the Step-Deviation method:

Daily Sales (₹ '000) Midpoint (m) Frequency (f) d' = (m - 35) / 10 f * d'

10 - 20 15 6 (15 - 35)/10 = -2 -12

20 - 30 25 14 (25 - 35)/10 = -1 -14

30 - 40 (Assumed A=35) 35 18 (35 - 35)/10 = 0 0

40 - 50 45 8 (45 - 35)/10 = +1 +8

50 - 60 55 4 (55 - 35)/10 = +2 +8

Total -- N = 50 -- SUM (fd') = -10

X_bar = A + [ SUM(fd') / N ] * c = 35 + [ (-10) / 50 ] * 10 = 35 + (-0.2 * 10) = 35 - 2 = ₹33,000.

Daily Sales (₹ '000)	Midpoint (m)	Frequency (f)	d' = (m - 35) / 10	f * d'
10 - 20	15	6	(15 - 35)/10 = -2	-12
20 - 30	25	14	(25 - 35)/10 = -1	-14
30 - 40 (Assumed A=35)	35	18	(35 - 35)/10 = 0	0
40 - 50	45	8	(45 - 35)/10 = +1	+8
50 - 60	55	4	(55 - 35)/10 = +2	+8
Total	--	N = 50	--	SUM (f*d') = -10

Part B (Correcting an Erroneous Mean): The average mark of 100 commerce students was found to be 64. Later, it was discovered that two marks originally recorded as 82 and 46 were misread as 28 and 64 respectively. Find the correct mean:

Original Incorrect Total = N * X_bar = 100 * 64 = 6,400.

Correct Total = Incorrect Total - (Incorrect Values) + (Correct Values) = 6400 - (28 + 64) + (82 + 46) = 6400 - 92 + 128 = 6,436.

Correct Mean = 6436 / 100 = 64.36 Marks.

12.3 Geometric Mean (GM) and Harmonic Mean (HM)

Geometric Mean (GM)

The Geometric Mean of (n) positive values is the (n)-th root of their product:

GM = sqrt[n]{X_1 cdot X_2 cdots X_n} = ext{Antilog}left( rac{sum log X}{n} ight)

Grouped Series: (GM = Antilogleft(∑ f log m) / (N)right).
Business Utility: The only mathematically correct average for calculating compound annual growth rates (CAGR), inflation indices, index numbers, and compound returns on financial assets.
Constraint: Fails if any value is zero or negative.

Harmonic Mean (HM)

The Harmonic Mean is the reciprocal of the arithmetic mean of the reciprocals of the individual values:

HM = rac{n}{sum left( rac{1}{X} ight)} = rac{N}{sum left( rac{f}{m} ight)}

Business Utility: Essential for computing average rates and speeds when the distance or expenditure is constant (e.g., dollar-cost averaging in equity mutual funds, fuel efficiency over fixed itineraries).
Weighting Property: Gives substantial weight to smaller observations and least weight to huge outliers.

12.4 The Median and Positional Values (Quartiles, Deciles, Percentiles)

The Median ((M)) is the positional average representing the exact middle observation when data is arranged in ascending or descending array. It splits the distribution into two equal halves (50% below, 50% above). Unlike the mean, the median is immune to extreme outliers, making it the preferred metric for skewed socioeconomic metrics such as executive compensation and household income.

Continuous Grouped Formulas for Positional Values PARTITION ALGEBRA

P a r t i t i o nV a l u e = L + [(R ank - c f_{p} r e) / f] * h

Rank Mapping Table:

Median (M): Rank = (N / 2)
First Quartile (Q_1): Rank = (N / 4) (25% boundary)
Third Quartile (Q_3): Rank = (3N / 4) (75% boundary)
k-th Decile (D_k): Rank = ((k cdot N) / 10) (where (k = 1, 2, dots, 9))
k-th Percentile (P_k): Rank = ((k cdot N) / 100) (where (k = 1, 2, dots, 99))

Where L is lower boundary of the respective class, cf_pre is cumulative frequency of the preceding class, f is class frequency, and h is class width.

12.5 The Mode and Empirical Relationship

The Mode ((Z)) is the value that occurs with the highest frequency in a distribution. In discrete series with multiple peaks or irregular clustering, the mode is determined using Grouping and Analysis Tables (forming pairs and triplets of column frequencies). In moderately asymmetrical unimodal distributions, Karl Pearson established the empirical relationship:

Mode = 3 * Median - 2 * Mean | Mean - Mode = 3 * ( Mean - Median )

Unit 13: Measures of Dispersion (Variation and Risk)

13.1 The Concept and Need for Measuring Dispersion

A measure of central tendency provides only a single central anchor. Two entirely different investment portfolios can have the exact same expected return of 12% per annum, yet one might fluctuate between 11% and 13% (low risk), while the other swings violently between -30% and +54% (high risk). Dispersion measures the degree of scatter, spread, or variability of observations around their central value.

Measures of dispersion are categorized into two structural types:

Absolute Measures: Expressed in the exact same physical units as the original data (e.g., ₹, kilograms, hours). Used to assess spread within a single dataset (Range, Quartile Deviation, Mean Deviation, Standard Deviation).
Relative Measures (Coefficients of Dispersion): Pure dimensionless ratios or percentages. Used to compare variation between two completely different series having different units or differing means (e.g., comparing wage variation in ₹ with worker age variation in years).

13.2 Range and Quartile Deviation

1. Range

The simplest measure of dispersion, defined as the difference between the maximum and minimum observations:

Range = L - S

Coefficient of Range = \frac{L - S}{L + S}

Business Utility: Daily stock market high/low price spreads; quality control control charts ( $X ˉ$ and (R) charts); meteorological weather forecasts.
Limitation: Completely ignores all intermediate data and is hyper-sensitive to extreme outliers.

2. Quartile Deviation (Semi-Interquartile Range)

Based on the distance between the third quartile ((Q_3)) and first quartile ((Q_1)), isolating the central 50% of the dataset:

Q D = \frac{Q _{3} - Q _{1}}{2}

Coefficient of Q D = \frac{Q _{3} - Q _{1}}{Q _{3} + Q _{1}}

Business Utility: Ideal for open-ended frequency distributions (e.g., "below ₹10,000" or "above ₹100,000") where extreme limits are unknown.

13.3 Mean Deviation (Average Deviation)

Mean Deviation ((MD)) is the arithmetic average of the absolute deviations of observations taken from a measure of central tendency (usually the Median or Mean), ignoring algebraic positive and negative signs:

Mean Deviation Mathematical Formulations ABSOLUTE SPREAD

MD = rac{sum f cdot |m - A|}{N}, quad ext{Coefficient of } MD = rac{MD}{A}

Where:

A = Central value chosen (Mean or Median). Note: Mathematically, Mean Deviation is at its absolute minimum when deviations are taken from the Median!
(|m - A|) = Absolute deviation with negative signs treated as positive.

13.4 Standard Deviation ((sigma)) and Variance ((sigma^2))

Introduced by Karl Pearson in 1893, the Standard Deviation ((sigma)) is the most mathematically rigorous, robust, and universally applied measure of dispersion. It is defined as the positive square root of the arithmetic mean of the squared deviations taken strictly from the arithmetic mean. Its square ((sigma^2)) is termed the Variance.

Standard Deviation Formulations across Series FOUNDATIONAL RISK METRIC

D i r ec t : s i g ma = S QR T [(S U M f * (m - X_{b} a r)^{2}) / N] ∣ S t e p - D e v : s i g ma = c * S QR T [(S U M f * d^{'2} / N) - (S U M f * d^{'} / N)^{2}]

Crucial Mathematical Properties of Standard Deviation:

Combined Standard Deviation ((sigma_{12})): For two pooled groups:

$s i g m a_{12} = \frac{N _{1} ( s i g m a _{1}^{2} + d _{1}^{2} ) + N _{2} ( s i g m a _{2}^{2} + d _{2}^{2} )}{N _{1} + N _{2}}$

Where $d_{1} = X ˉ_{1} - X ˉ_{12}$ and $d_{2} = X ˉ_{2} - X ˉ_{12}$ .
Normal Distribution Empirical Rule: For symmetrical bell-shaped distributions:
- $X ˉ \pm 1 s i g ma$ encompasses exactly 68.27% of all items.
- $X ˉ \pm 2 s i g ma$ encompasses exactly 95.45% of all items.
- $X ˉ \pm 3 s i g ma$ encompasses exactly 99.73% of all items (foundation of Six Sigma quality control).
Invariance: Standard deviation is independent of change of origin (adding or subtracting a constant does not change (sigma)), but is dependent on change of scale (multiplying by a constant scales (sigma) proportionally).

13.5 Coefficient of Variation (CV) & Commercial Risk Comparison

The Coefficient of Variation ((CV)), developed by Karl Pearson, expresses the standard deviation as a percentage of the arithmetic mean:

C oe f f i c i e n t o f V a r ia t i o n (C V) = (s i g ma / X_{b} a r) * 100%

In corporate finance, equity portfolio management, and HR productivity comparisons:

A series with a lower CV is deemed more consistent, stable, uniform, and reliable.
A series with a higher CV exhibits greater variability, dispersion, risk, and volatility.

Numerical Problem 13.1: Financial Risk Assessment of Two Competing Stocks
An investment manager evaluates the closing share prices of two blue-chip corporations over a 10-day trading window:

Stock Alpha: Mean price $X ˉ_{A} = Rs. 450$ ; Standard deviation (sigma_A = ₹18.0).

Stock Beta: Mean price $X ˉ_{B} = Rs. 1, 200$ ; Standard deviation (sigma_B = ₹36.0).

Stock Alpha CV = ( sigma_A / X_bar_A ) * 100 = ( 18.0 / 450 ) * 100 = 4.00%.

Stock Beta CV = ( sigma_B / X_bar_B ) * 100 = ( 36.0 / 1200 ) * 100 = 3.00%.

Financial Evaluation: Although Stock Beta has a higher absolute standard deviation (₹36 vs ₹18), its Coefficient of Variation is lower (3.0% vs 4.0%). Therefore, relative to its price level, Stock Beta is more stable and less volatile, making it the superior defensive equity investment!

Unit 14: Skewness and Kurtosis: Distributional Geometry

14.1 The Concept of Skewness (Asymmetry)

While central tendency indicates the center and dispersion measures the spread, neither reveals the direction of deviation. Two distributions may possess identical means (e.g., 50) and identical standard deviations (e.g., 10), yet one may have data evenly balanced around the center, while the other exhibits extreme values concentrated far to the right. Skewness measures the lack of symmetry in a statistical distribution.

Characteristic	Symmetrical Distribution	Positively Skewed Distribution	Negatively Skewed Distribution
Curve Geometry	Perfect bell shape; left and right halves are identical mirror images.	The elongated tail of the distribution stretches toward the right (higher values).	The elongated tail of the distribution stretches toward the left (lower values).
Order of Averages	$Mean = Median = Mode$	$Mean > Median > Mode$	$Mean < Median < Mode$
Quartile Distances	(Q_3 - M = M - Q_1)	((Q_3 - M) > (M - Q_1))	((Q_3 - M) < (M - Q_1))
Commercial Examples	Machinery component dimensions in high-precision manufacturing.	Corporate income distribution, executive salaries, insurance claim sizes.	Retirement ages, lifespan of premium electronic hardware components.

14.2 Mathematical Measures of Skewness

Statisticians compute relative coefficients of skewness to quantify asymmetry independently of data units:

Classical Coefficients of Skewness ASYMMETRY METRICS

Sk_p = rac{ ext{Mean} - ext{Mode}}{sigma}, quad Sk_b = rac{(Q_3 - M) - (M - Q_1)}{Q_3 - Q_1}

Interpretation Parameters:

Karl Pearson's Coefficient ((Sk_p)): If the mode is ill-defined, deploy (Sk_p = (3 $Mean - Median$ ) / (sigma)). The value normally ranges between (-3) and (+3). A value of 0 indicates perfect symmetry.
Bowley's Quartile Coefficient ((Sk_b)): Simplifies to:

$S k_{b} = \frac{Q _{3} + Q _{1} - 2 M}{Q _{3} - Q _{1}}$

Bound strictly between (-1) and (+1). Ideal for open-ended frequency tables where mean and SD cannot be computed.
Kelly's Measure ((Sk_k)): Based on the 10th and 90th percentiles: $S k_{k} = \frac{P _{90} + P _{10} - 2 M}{P _{90} - P _{10}}$ .

14.3 The Concept and Classification of Kurtosis (Peakedness)

Kurtosis (derived from the Greek word kyrtos, meaning bulging or convex) measures the relative peakedness or flatness of a frequency distribution curve compared to the standard Normal (Gaussian) distribution. It indicates the degree of data concentration around the central modal peak versus the heavy-tailed dispersion.

1. Leptokurtic (Slender / Sharp)

A distribution that has a higher, sharper central peak and fatter tails than the normal curve.

Moment Ratio: $β_{2} > 3$ (or Excess Kurtosis (gamma_2 > 0)).
Business Meaning: High concentration around a single value; higher probability of extreme outlier shocks (fat tails).

2. Mesokurtic (Normal)

The standard baseline bell-shaped Gaussian curve.

Moment Ratio: $β_{2} = 3$ (or Excess Kurtosis $g amm a_{2} = β_{2} - 3 = 0$ ).
Business Meaning: Predictable, random Gaussian error distribution matching classical central limit theorem expectations.

3. Platykurtic (Broad / Flat)

A distribution with a broad, flat-topped peak and thinner tails than the normal curve.

Moment Ratio: $β_{2} < 3$ (or Excess Kurtosis (gamma_2 < 0)).
Business Meaning: Wide, even dispersion with minimal heavy-tail outlier events.

Central Moments and Kurtosis Mathematical Ratios MOMENTS ALGEBRA

m u_{2} = s i g m a^{2} ∣ m u_{4} = [S U M f * (m - X_{b} a r)^{4}] / N ∣ b e t a_{2} = m u_{4} / (m u_{2})^{2}

Where (mu_2) is the second central moment (variance) and (mu_4) is the fourth central moment. Excess Kurtosis is defined as:

g amm a_{2} = β_{2} - 3

Numerical Problem 14.1: Comprehensive Asymmetry and Quartile Skewness Calculation
In an analysis of monthly revenue across 250 franchise stores, the partition metrics were computed as: First Quartile $Q_{1} = Rs. 42.0 Lakhs$ ; Median $M = Rs. 55.0 Lakhs$ ; Third Quartile $Q_{3} = Rs. 78.0 Lakhs$ . In addition, Mean $X ˉ = Rs. 58.5 Lakhs$ and Standard Deviation $s i g ma = Rs. 15.0 Lakhs$ .

Step 1: Compute Bowley's Skewness: Sk_b = [ Q_3 + Q_1 - 2*M ] / [ Q_3 - Q_1 ]

Numerator = 78.0 + 42.0 - 2*(55.0) = 120.0 - 110.0 = +10.0.

Denominator = 78.0 - 42.0 = 36.0.

Sk_b = +10.0 / 36.0 = +0.2778 (Moderately Positively Skewed).

Step 2: Compute Karl Pearson's Skewness: Sk_p = 3*(Mean - Median) / sigma

Sk_p = 3*( 58.5 - 55.0 ) / 15.0 = 3*( 3.5 ) / 15.0 = 10.5 / 15.0 = +0.7000 (Positively Skewed).

Business Conclusion: Both coefficients confirm positive skewness ( $Mean > Median$ ). A cluster of highly successful flagship stores pulls the revenue tail to the right, while the majority of franchisees operate below the average revenue line!

14.4 Summary Matrix of Univariate Descriptive Statistics

Statistical Dimension	Primary Metrics	Key Mathematical Characteristic	Core Business Application
Center (Location)	Arithmetic Mean, Median, Mode, GM, HM	(sum $X - X ˉ$ = 0); Median splits data 50/50; Mode = peak frequency.	Corporate benchmark targets, baseline costs, expected sales volumes.
Spread (Dispersion)	Standard Deviation, Variance, QD, Range, CV	$s i g ma = Variance$ ; (CV = $s i g ma / X ˉ$ × 100).	Risk evaluation, volatility modeling, consistency checks, Six Sigma quality control.
Symmetry (Skewness)	Karl Pearson's ((Sk_p)), Bowley's ((Sk_b))	$S k_{p} = \frac{X ˉ - Z}{s i g ma}$ ; $S k_{b} = \frac{Q _{3} + Q _{1} - 2 M}{Q _{3} - Q _{1}}$ .	Income disparity analysis, insurance risk tail sizing, customer spending patterns.
Peakedness (Kurtosis)	$β_{2} = m u_{4} / m u_{2}^{2}$ , Excess Kurtosis (gamma_2)	Mesokurtic ( $β_{2} = 3$ ); Leptokurtic ( $β_{2} > 3$ ); Platykurtic ( $β_{2} < 3$ ).	Financial tail risk, black swan vulnerability, asset price crash probabilities.