Central Tendency, Dispersion and Skewness
UGC NET Commerce: mean, median, mode, GM and HM; range, QD, MD, SD and CV; Karl Pearson and Bowley skewness, moments and kurtosis, with worked examples.
- 21 min
On this page · 13 sections
Prerequisites
- Fractions, Ratio, Percentage and Averages — the simple and weighted average you already use in Paper 1.
- Data Interpretation Calculations — percentage change and growth rates, which lead to the geometric mean.
- Business Statistics and Research Methods: chapter overview — where this page sits among the eleven official lines.
The Hook: One Factory, Three “Average” Wages
1. Measures of Central Tendency
An average is a single value that represents the whole distribution. The textbook requisites of a good average are that it should be rigidly defined, based on all observations, easy to understand and compute, capable of further algebraic treatment, least affected by sampling fluctuations and not unduly affected by extreme values. No single average meets all six, which is why there are five of them.
Arithmetic mean (AM)
| Data | Formula |
|---|---|
| Ungrouped | \( \bar{X} = \dfrac{\sum X}{n} \) |
| Discrete or grouped (direct) | \( \bar{X} = \dfrac{\sum fX}{N} \), with \( X \) the mid-value of each class |
| Short-cut (assumed mean A) | \( \bar{X} = A + \dfrac{\sum fd}{N} \), with \( d = X - A \) |
| Step-deviation | \( \bar{X} = A + \dfrac{\sum fd'}{N} \times h \), with \( d' = (X - A)/h \) |
| Weighted | \( \bar{X}_w = \dfrac{\sum wX}{\sum w} \) |
| Combined mean of two groups | \( \bar{X}_{12} = \dfrac{n_1 \bar{X}_1 + n_2 \bar{X}_2}{n_1 + n_2} \) |
Mathematical properties that turn up as statements:
- The sum of deviations from the mean is zero: \( \sum (X - \bar{X}) = 0 \).
- The sum of squared deviations is least when taken from the mean.
- Adding a constant to every value adds it to the mean; multiplying every value by a constant multiplies the mean by it.
- \( \sum X = n\bar{X} \), so a wrongly copied value can be corrected without recomputing everything: new mean \( = (n\bar{X} - \text{wrong} + \text{correct})/n \).
Merits: rigidly defined, uses all values, algebraically tractable, stable across samples. Demerits: distorted by extreme values; cannot be found for open-end classes without assuming limits; can be a value that does not exist (2.4 children per family).
Geometric mean (GM)
$$GM = \sqrt[n]{X_1 \times X_2 \times \cdots \times X_n}, \qquad \log GM = \frac{\sum \log X}{n}$$Use it for rates of change, ratios and percentages that compound: average growth rate of sales, average rate of return over several years, and the construction of index numbers (Fisher’s ideal index is the GM of the Laspeyres and Paasche indices). It is undefined if any value is zero or negative.
Harmonic mean (HM)
$$HM = \frac{n}{\sum \left( 1/X \right)}$$Use it for rates expressed per unit when the numerator quantity is fixed: average speed over equal distances, average price when equal amounts of money are spent on each purchase.
Relation among the three means. For positive values that are not all equal, \( AM > GM > HM \); they are equal only when every value is the same. For two values, \( GM^2 = AM \times HM \). For 4 and 16: AM = 10, GM = 8, HM = 6.4, and \( 8^2 = 10 \times 6.4 \).
Median and partition values
The median is the middle value of the data arranged in order. It divides the distribution into two equal halves.
- Ungrouped: the \( \left(\frac{n+1}{2}\right) \)th value.
- Grouped (continuous classes):
where \( L \) is the lower limit of the median class, \( cf \) the cumulative frequency of the class before it, \( f \) the frequency of the median class and \( h \) its width.
The same interpolation gives the partition values: quartiles (\( Q_1 \) at \( N/4 \), \( Q_3 \) at \( 3N/4 \)), deciles (\( D_k \) at \( kN/10 \)) and percentiles (\( P_k \) at \( kN/100 \)). Note \( Q_2 = D_5 = P_{50} = \) median.
Properties: the sum of absolute deviations is least when taken from the median; it is unaffected by extreme values and can be found for open-end distributions; it suits qualitative or ranked data (intelligence, honesty). It can be located graphically as the X-coordinate where the less-than and more-than ogives intersect.
Mode
The mode is the value that occurs most often: the point of maximum concentration. For a grouped distribution:
$$\text{Mode} = L + \frac{f_1 - f_0}{2f_1 - f_0 - f_2} \times h$$with \( f_1 \) the frequency of the modal class, \( f_0 \) of the class before it and \( f_2 \) of the class after it. If the frequencies are irregular, a grouping table is used to locate the modal class. Graphically, the mode is read from the histogram.
A distribution can be unimodal, bimodal or multimodal; the mode is then ill-defined, which is its main weakness. It is the right average for “most common size of shoe or shirt to stock”.
Empirical relation (Karl Pearson), for moderately skewed distributions:
$$\text{Mode} = 3\,\text{Median} - 2\,\text{Mean}$$If mean = 40 and median = 38, mode \( \approx 3(38) - 2(40) = 34 \).
Which average for which situation
| Situation | Best average | Why |
|---|---|---|
| Average marks, output, income for further algebra | Arithmetic mean | Uses all values, algebraically tractable |
| Average growth rate, rate of return, index numbers | Geometric mean | Compounding; gives less weight to large values |
| Average speed over equal distances; average price for equal outlays | Harmonic mean | Rates per unit with fixed numerator |
| Income distribution with a few very rich; open-end classes; ranked qualities | Median | Not affected by extremes |
| Most popular size, design or brand | Mode | Point of maximum concentration |
2. A Full Worked Example (Grouped Data)
The daily wages (₹ hundred) of 50 workers:
| Wages | 0–10 | 10–20 | 20–30 | 30–40 | 40–50 |
|---|---|---|---|---|---|
| Workers (f) | 5 | 20 | 15 | 5 | 5 |
| Mid-value (X) | 5 | 15 | 25 | 35 | 45 |
| \( d' = (X - 25)/10 \) | −2 | −1 | 0 | 1 | 2 |
| \( fd' \) | −10 | −20 | 0 | 5 | 10 |
| \( fd'^2 \) | 20 | 20 | 0 | 5 | 20 |
| Cumulative f | 5 | 25 | 40 | 45 | 50 |
We continue with this table for dispersion and skewness below.
3. Measures of Dispersion
Dispersion measures the scatter of values around the average. Two distributions can share a mean of 50 and yet one ranges from 48 to 52 and the other from 0 to 100.
Absolute measures are in the units of the data (rupees, kg). Relative measures (coefficients) are pure numbers, used to compare series in different units or with different means.
| Measure | Absolute | Relative (coefficient) |
|---|---|---|
| Range | \( L - S \) (largest minus smallest) | \( \dfrac{L - S}{L + S} \) |
| Quartile deviation (semi-interquartile range) | \( \dfrac{Q_3 - Q_1}{2} \) | \( \dfrac{Q_3 - Q_1}{Q_3 + Q_1} \) |
| Mean deviation | \( \dfrac{\sum f \lvert X - A \rvert}{N} \), A = mean or median | \( \dfrac{MD}{A} \) |
| Standard deviation | \( \sigma = \sqrt{\dfrac{\sum f (X - \bar{X})^2}{N}} \) | Coefficient of SD \( = \sigma / \bar{X} \) |
| Coefficient of variation | — | \( CV = \dfrac{\sigma}{\bar{X}} \times 100 \) |
Range and quartile deviation
The range is the quickest measure and is used in statistical quality control (control charts for range) and for daily price movements, but it depends only on two extreme values. The quartile deviation ignores the top and bottom 25% of values, so it is not affected by extremes and can be computed for open-end classes; it ignores half the data.
Mean deviation
Mean deviation is the average of absolute deviations. Taken from the median, it is the smallest possible (property of the median). Ignoring signs makes it unsuitable for further algebra, which is why the standard deviation replaced it.
Standard deviation and variance
The standard deviation (the term was introduced by Karl Pearson) is the root mean square deviation from the mean. Variance is \( \sigma^2 \). Working formulas:
$$\sigma = \sqrt{\frac{\sum fX^2}{N} - \left(\frac{\sum fX}{N}\right)^2}, \qquad \sigma = h \sqrt{\frac{\sum fd'^2}{N} - \left(\frac{\sum fd'}{N}\right)^2}$$Properties:
- SD is independent of change of origin (adding 5 to every value leaves it unchanged) but not of scale (multiplying every value by 3 multiplies SD by 3 and variance by 9).
- SD is the minimum root mean square deviation: root mean square deviation about any other point is larger.
- For two values a and b, \( \sigma = \lvert a - b \rvert / 2 \). For the first n natural numbers, \( \sigma = \sqrt{(n^2 - 1)/12} \).
- Combined SD of two groups:
- For a normal distribution, approximately: QD \( \approx \frac{2}{3}\sigma \) and MD \( \approx \frac{4}{5}\sigma \), so \( 6\,QD \approx 5\,MD \approx 4\,\sigma \) (more exactly QD \( \approx 0.6745\sigma \), MD \( \approx 0.7979\sigma \)).
Coefficient of variation
The CV (Karl Pearson) expresses SD as a percentage of the mean. Lower CV = more consistent, uniform, stable, homogeneous; higher CV = more variable. It is the measure for comparing two batsmen, two shares’ returns, or two machines’ output when the means differ.
Lorenz curve
The Lorenz curve (Max O. Lorenz, 1905) is a graphic measure of dispersion: cumulative percentage of a variable (income, wealth, sales) plotted against the cumulative percentage of units (persons, firms). The diagonal is the line of equal distribution; the farther the curve sags from it, the greater the inequality. The Gini coefficient is the area between the diagonal and the curve divided by the whole area under the diagonal: 0 for perfect equality, 1 for perfect inequality.
4. Measures of Skewness
Skewness means lack of symmetry. Dispersion tells you how much the values spread; skewness tells you in which direction the spread is lopsided.
| Distribution | Relation of averages | Tail | Quartiles |
|---|---|---|---|
| Symmetrical | Mean = Median = Mode | Equal on both sides | \( Q_3 - M = M - Q_1 \) |
| Positively skewed (right-skewed) | Mean > Median > Mode | Longer to the right | \( Q_3 - M > M - Q_1 \) |
| Negatively skewed (left-skewed) | Mean < Median < Mode | Longer to the left | \( Q_3 - M < M - Q_1 \) |
The median always lies between the mean and the mode, and in a moderately skewed distribution the mean–median distance is about one-third of the mean–mode distance (which is just the empirical relation rearranged).
Coefficients of skewness
| Coefficient | Formula | Range | Based on |
|---|---|---|---|
| Karl Pearson (first) | \( Sk_P = \dfrac{\text{Mean} - \text{Mode}}{\sigma} \) | No fixed limit; usually within ±1 | Mean, mode, SD |
| Karl Pearson (second) | \( Sk_P = \dfrac{3(\text{Mean} - \text{Median})}{\sigma} \) | −3 to +3 | Used when mode is ill-defined |
| Bowley (quartile) | \( Sk_B = \dfrac{Q_3 + Q_1 - 2\,\text{Median}}{Q_3 - Q_1} \) | −1 to +1 | Middle 50%; good for open-end classes |
| Kelly (percentile) | \( Sk_K = \dfrac{P_{90} + P_{10} - 2P_{50}}{P_{90} - P_{10}} \) | −1 to +1 | Middle 80% |
| Moment-based | \( \beta_1 = \dfrac{\mu_3^2}{\mu_2^3} \), \( \gamma_1 = \dfrac{\mu_3}{\sigma^3} \) | — | Third central moment; \( \beta_1 \) gives size only, sign comes from \( \mu_3 \) |
Moments and kurtosis
Central moments are averages of powers of deviations from the mean: \( \mu_r = \frac{1}{N}\sum f(X - \bar{X})^r \). Always \( \mu_0 = 1 \), \( \mu_1 = 0 \), and \( \mu_2 = \sigma^2 \). From raw moments \( \mu_r' \) about an arbitrary origin:
$$\mu_2 = \mu_2' - \mu_1'^2, \qquad \mu_3 = \mu_3' - 3\mu_2'\mu_1' + 2\mu_1'^3$$| Moment | Describes |
|---|---|
| First raw moment | Mean (central tendency) |
| Second central | Variance (dispersion) |
| Third central | Skewness (direction and size of asymmetry) |
| Fourth central | Kurtosis (peakedness, tail weight) |
Kurtosis is measured by \( \beta_2 = \mu_4/\mu_2^2 \) (and excess kurtosis \( \gamma_2 = \beta_2 - 3 \)):
| Type | \( \beta_2 \) | \( \gamma_2 \) | Shape |
|---|---|---|---|
| Leptokurtic | > 3 | > 0 | More peaked, heavier tails |
| Mesokurtic | = 3 | = 0 | Normal curve |
| Platykurtic | < 3 | < 0 | Flatter |
Kurtosis is not named in the official line, but it sits next to skewness in every textbook and in statement-based questions (“a normal curve is mesokurtic”).
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| Sum of deviations from mean = 0 | vs | Sum of squared deviations least from mean; sum of absolute deviations least from median |
| GM: growth rates, index numbers | vs | HM: speeds over equal distances, equal outlays |
| Absolute dispersion (range, QD, MD, SD): in data units | vs | Relative dispersion (coefficients, CV): unit-free, for comparison |
| Dispersion: how much the values spread | vs | Skewness: which way the spread is lopsided; kurtosis: how peaked |
| Pearson’s coefficient: uses mean, mode or median and SD; ±3 | vs | Bowley’s coefficient: uses quartiles; ±1 |
| Ogive (cumulative frequency curve): locates median, quartiles | vs | Histogram: locates the mode |
| Leptokurtic: \( \beta_2 > 3 \), peaked | vs | Platykurtic: \( \beta_2 < 3 \), flat |
| Lower CV: more consistent | vs | Higher SD alone: says nothing about consistency if the means differ |
Common Mistakes
Practice
All questions below are original practice questions written in the UGC NET format. They are not previous-year questions.
Level 1: Foundation
- Which average is most suitable for finding the average rate of growth of a company’s profits over five years?
- For two positive numbers, AM = 25 and GM = 20. Find the HM.
- What are the limits of Bowley’s coefficient of skewness?
Answers
- Geometric mean: growth compounds.
- \( HM = GM^2/AM = 400/25 = \) 16.
- −1 to +1.
Level 2: Exam-style MCQs
Q1. (Practice) In a moderately skewed distribution, the mean is 52 and the median is 50. The approximate mode is:
- (a) 46
- (b) 48
- (c) 54
- (d) 56
Answer and explanation
(a) 46. Mode \( = 3 \times 50 - 2 \times 52 = 150 - 104 = 46 \). Mean > median > mode, so the distribution is positively skewed, as expected.
Q2. (Practice) Each observation of a series is multiplied by 4 and then 10 is added to it. If the original variance was 9, the new standard deviation is:
- (a) 12
- (b) 22
- (c) 36
- (d) 144
Answer and explanation
(a) 12. Original SD = 3. Adding 10 (origin) has no effect; multiplying by 4 (scale) makes SD \( 4 \times 3 = 12 \). The new variance is 144, which is option (d), the trap.
Q3. (Practice) Two machines fill packets. Machine P: mean 500 g, SD 10 g. Machine Q: mean 250 g, SD 6 g. Which statement is correct?
- (a) P is more consistent because its CV is 2%
- (b) Q is more consistent because its SD is smaller
- (c) Q is more consistent because its CV is 2.4%
- (d) Both are equally consistent
Answer and explanation
(a). CV of P \( = 10/500 \times 100 = 2\% \); CV of Q \( = 6/250 \times 100 = 2.4\% \). P has the lower CV and is more consistent even though its SD is larger. Option (b) is the classic SD trap.
Q4. (Practice) Consider the following statements:
- The sum of absolute deviations is minimum when taken from the median.
- The standard deviation is independent of change of origin.
- Karl Pearson’s coefficient of skewness based on the median lies between −1 and +1.
Which of the statements given above is/are correct?
- (a) 1 and 2 only
- (b) 2 and 3 only
- (c) 1 and 3 only
- (d) 1, 2 and 3
Answer and explanation
(a) 1 and 2 only. Statement 3 is wrong: the median-based Pearson coefficient \( 3(\text{Mean} - \text{Median})/\sigma \) lies between −3 and +3. The ±1 limit belongs to Bowley’s coefficient.
Q5. (Practice) For a distribution, \( Q_1 = 30 \), median = 40 and \( Q_3 = 60 \). Bowley’s coefficient of skewness is:
- (a) 0.25
- (b) 0.33
- (c) −0.33
- (d) 0.50
Answer and explanation
(b) 0.33. \( (60 + 30 - 80)/(60 - 30) = 10/30 \approx 0.33 \). Positive: the upper quartile is farther from the median than the lower quartile.
Level 3: Assertion–reason, match and sequence
For assertion–reason questions, use these options:
- (a) Both A and R are true, and R is the correct explanation of A
- (b) Both A and R are true, but R is not the correct explanation of A
- (c) A is true, but R is false
- (d) A is false, but R is true
Q1. (Practice) Assertion (A): The median is a more appropriate average than the arithmetic mean for the distribution of household wealth. Reason (R): The median is not affected by extreme values.
Answer and explanation
(a). Wealth is heavily positively skewed by a few very large holdings. The median ignores their size, which is exactly why it represents the typical household better.
Q2. (Practice) Assertion (A): The coefficient of variation is used to compare the variability of two series expressed in different units. Reason (R): The standard deviation is independent of change of scale.
Answer and explanation
(c). A is true: the CV is a unit-free ratio. R is false: SD is independent of change of origin, not of scale.
Q3. (Practice) Match List I with List II.
| List I (Situation) | List II (Appropriate measure) |
|---|---|
| A. Average speed over equal distances | I. Mode |
| B. Most demanded shoe size | II. Geometric mean |
| C. Average rate of population growth | III. Harmonic mean |
| D. Comparing consistency of two share prices | IV. Coefficient of variation |
- (1) A-III, B-I, C-II, D-IV
- (2) A-II, B-I, C-III, D-IV
- (3) A-III, B-IV, C-II, D-I
- (4) A-I, B-III, C-II, D-IV
Answer and explanation
(1) A-III, B-I, C-II, D-IV. HM for rates over a fixed numerator, mode for the most frequent value, GM for compounding growth, CV for relative consistency.
Q4. (Practice) In a positively skewed distribution, arrange the following in increasing order of value:
A. Mean
B. Mode
C. Median
(1) A, C, B
(2) B, C, A
(3) C, B, A
(4) B, A, C
Answer and explanation
(2) B, C, A. Mode < median < mean when the tail is on the right.
Q5. (Practice) Arrange the following measures for the same set of positive, unequal values in decreasing order:
A. Harmonic mean
B. Arithmetic mean
C. Geometric mean
(1) B, C, A
(2) C, B, A
(3) A, C, B
(4) B, A, C
Answer and explanation
(1) B, C, A. AM > GM > HM whenever the values are positive and not all equal.
Quick Revision Table
| Item | Formula or fact |
|---|---|
| AM (step deviation) | \( A + \frac{\sum fd'}{N} h \) |
| Combined mean | \( \frac{n_1\bar{X}_1 + n_2\bar{X}_2}{n_1 + n_2} \) |
| GM | nth root of product; growth rates, index numbers |
| HM | \( n / \sum(1/X) \); speeds, equal outlays |
| AM, GM, HM | \( AM \ge GM \ge HM \); two values: \( GM^2 = AM \times HM \) |
| Median (grouped) | \( L + \frac{N/2 - cf}{f} h \); ogives intersect at it |
| Mode (grouped) | \( L + \frac{f_1 - f_0}{2f_1 - f_0 - f_2} h \); histogram |
| Empirical relation | Mode = 3 Median − 2 Mean |
| Range / QD | \( L - S \) / \( (Q_3 - Q_1)/2 \) |
| SD | Root mean square deviation from mean; origin-free, scale-dependent |
| CV | \( \sigma/\bar{X} \times 100 \); lower = more consistent |
| Normal: QD, MD, SD | \( \approx \frac{2}{3}\sigma \), \( \frac{4}{5}\sigma \), \( \sigma \) |
| Lorenz curve | Graphic inequality; Gini coefficient 0 to 1 |
| Skewness | Positive: mean > median > mode |
| Pearson / Bowley / Kelly | Mean–mode or 3(mean–median) over SD (±3) / quartiles (±1) / percentiles (±1) |
| Kurtosis | \( \beta_2 = \mu_4/\mu_2^2 \); 3 meso, > 3 lepto, < 3 platy |
Key Takeaways
What’s Next
- Correlation and Regression — from one variable to the relationship between two
- Probability and Probability Distributions — the normal curve, whose mean, SD and kurtosis you have just met
- Financial Statement Analysis — averages and ratios in accounting
- Quick Revision: Business Statistics and Research Methods