Measures of dispersion

Two students can average the exact same marks and still be completely different learners — one rock-steady, the other wildly unpredictable. The mean hides this; it collapses a whole data set into a single number and throws away every trace of how spread out the values actually are. Measures of dispersion are the fix: a family of numbers — range, mean deviation, variance, standard deviation and the coefficient of variation — that quantify exactly how far, on average, data values scatter around their centre. This article opens the Statistics chapter of CBSE Class 11 Mathematics (Unit V), building directly on the mean, median and mode from Class 9–10. You will see why "same mean" never means "same behaviour," derive the variance and standard deviation formula from scratch, and work a fully quantified example comparing two data sets step by step.

The core insight

Dispersion measures — range, mean deviation, variance, standard deviation and the coefficient of variation — quantify how far data values scatter around a central value, revealing consistency differences that the mean alone conceals.

The core idea

Consider two students' marks (out of 10) across 5 tests: Student A scores 6, 6, 6, 6, 6; Student B scores 0, 2, 6, 10, 12. Both have mean $\bar x = 6$ , yet A is perfectly consistent and B is erratic. A single number that captures this difference is what dispersion measures provide.

Range is the crudest measure, using only the two extremes:

\text{Range} = x_{\max} - x_{\min}

where: $x_{\max}$ and $x_{\min}$ are the largest and smallest values in the data set. It is quick to compute but ignores every value in between.

Mean deviation averages the absolute distance of each value from a chosen centre $a$ (usually the mean $\bar x$ or median $M$ ):

\text{MD} = \frac{1}{n}\sum_{i=1}^{n} |x_i - a|

where: $n$ is the number of observations, $x_i$ is the $i$ -th value, and $a$ is the reference point. Absolute value is needed because deviations above and below the mean would otherwise cancel to zero.

Variance and standard deviation replace absolute value with squaring, which is more mathematically workable and gives more weight to values far from the mean:

\sigma^2 = \frac{1}{n}\sum_{i=1}^{n} (x_i - \bar x)^2, \qquad \sigma = \sqrt{\sigma^2}

where: $\sigma^2$ is the variance, $\sigma$ is the standard deviation (same units as the data), and $\bar x$ is the mean of all $n$ values.

Coefficient of variation turns standard deviation into a unit-free percentage, so two series with different means or different units can be compared fairly:

\text{CV} = \frac{\sigma}{\bar x} \times 100\%

A larger CV means relatively more variability, regardless of the units the original data were measured in. Range, mean deviation and standard deviation are absolute measures — they carry the same units as the data — so only CV should be used to compare series measured in different units or with very different means.

Derivation of the variance and standard deviation formula

Derivation AVariance and standard deviation from raw data

The goal is to build, from first principles, a single number that measures the average spread of a data set around its mean without the deviations above and below cancelling to zero, using only the arithmetic mean the student already knows how to compute. Start from the natural but broken idea of simply averaging the deviations, see why it fails, and repair it by squaring.

Step 1. For each observation $x_i$ in a data set of $n$ values with mean $\bar x = \frac{1}{n}\sum x_i$ , define the deviation $d_i = x_i - \bar x$ . Some deviations are positive (value above the mean) and some are negative (value below the mean).

Step 2. Try averaging the deviations directly: $\frac{1}{n}\sum d_i$ . By the defining property of the mean, $\sum (x_i - \bar x) = \sum x_i - n\bar x = n\bar x - n\bar x = 0$ always, for any data set. So this naive average is always exactly zero — it carries no information about spread at all, and cannot be used as a dispersion measure.

Step 3. Repair this by squaring each deviation before averaging, since $d_i^2 \geq 0$ removes the sign problem while keeping the deviation's size. Define $\sigma^2 = \frac{1}{n}\sum (x_i - \bar x)^2$ , the variance. Because every term is non-negative, the sum cannot cancel to zero (unless every value equals the mean), and larger individual deviations contribute disproportionately more due to the squaring.

Step 4. Variance is expressed in squared units of the original data (for example, cm $^2$ if the data are heights in cm), which is awkward to interpret. Taking the positive square root restores the original units: $\sigma = \sqrt{\sigma^2}$ , the standard deviation — the standard measure of spread used throughout statistics.

Worked example

Extend the two sections of Worked Example 1 from the introductory pass: Section X recorded heights 150, 152, 154, 156, 158 cm; Section Y recorded 140, 150, 155, 160, 170 cm. Both have nearly equal means, but the range calculation already hinted Section Y is far more spread out. Quantify this fully with variance, standard deviation and CV.

Step 1. Section X: mean $\bar x_X = 154$ . Deviations from the mean: $-4, -2, 0, 2, 4$ . Squared: $16, 4, 0, 4, 16$ , summing to $40$ . Variance $\sigma_X^2 = 40/5 = 8$ , so standard deviation $\sigma_X = \sqrt{8} \approx 2.83$ cm.

Step 2. Section Y: mean $\bar x_Y = 155$ . Deviations: $-15, -5, 0, 5, 15$ . Squared: $225, 25, 0, 25, 225$ , summing to $500$ . Variance $\sigma_Y^2 = 500/5 = 100$ , so standard deviation $\sigma_Y = \sqrt{100} = 10$ cm.

Step 3. Compute CV for a fair comparison: $\text{CV}_X = (2.83/154) \times 100 \approx 1.84\%$ , and $\text{CV}_Y = (10/155) \times 100 \approx 6.45\%$ .

Step 4. Interpret. Section Y's standard deviation ( $10$ cm) is more than three times Section X's ( $2.83$ cm), and its CV ( $6.45\%$ ) is likewise over three times larger, despite the two sections having nearly identical mean height. Section Y's heights are far more variable than Section X's — the earlier range comparison (30 cm vs. 8 cm) was correct in direction, and standard deviation now confirms and quantifies it precisely.

Common traps

Averaging raw deviations $(x_i - \bar x)$ directly instead of squaring — this sum is always zero and reveals nothing about spread.
Forgetting to take the square root after computing variance, and quoting $\sigma^2$ (wrong units) where $\sigma$ was asked for.
Comparing raw standard deviations of two series with different means or units — only the coefficient of variation is valid for that comparison.
Assuming a bigger range always means "more spread overall" — range uses only the two extreme values and can be misleading if one value is unusually far out.

Ask the tutor: Want to work through a mean-deviation or coefficient-of-variation calculation on a data set of your own?

Why do we square the deviations instead of just using their absolute values, like mean deviation does?

Both fix the same problem — deviations above and below the mean cancel to zero if averaged directly. Squaring is preferred over absolute value because it is smoother and easier to work with algebraically: it can be differentiated, it combines nicely with other squared quantities, and it underlies powerful results elsewhere in statistics. Mean deviation is simpler conceptually but variance and standard deviation are the measures built on for further theory, which is why the chapter develops both but standard deviation becomes the default.

If two data sets have the same standard deviation, does that mean they have the same spread?

Only if they also have similar means and are measured in the same units. Standard deviation is an absolute measure, so a $\sigma = 10$ on data centred around $1000$ represents a tiny relative spread, while the same $\sigma = 10$ on data centred around $20$ represents huge relative spread. This is exactly the situation the coefficient of variation is built to resolve — always check CV before declaring two data sets "equally variable."

Why is variance never negative, and what does a variance of zero mean?

Variance is a sum of squared terms divided by $n$ , and a square is never negative, so the sum — and hence the variance — cannot be negative. Variance equals exactly zero only when every single data value equals the mean, meaning there is no spread at all: a perfectly constant data set. Any nonzero variance signals genuine variability in the data, and if you ever compute a negative variance or standard deviation, it is a certain sign of an arithmetic mistake.

Measures of dispersion

The core idea

Derivation of the variance and standard deviation formula

Worked example

Common traps

Before this

Where this leads