Measures of Dispersion: Range, Variance, Standard Deviation & More

What Are Measures of Dispersion?

Measures of central tendency tell you where data clusters, but they leave out half the story. Two datasets can have the same mean yet look nothing alike:

Dispersion is also called variability, spread, or scatter. The larger the dispersion measure, the more spread out the data values are. A dispersion of zero means every value is identical.

Absolute vs. Relative Measures of Dispersion

All dispersion measures fall into one of two categories. Absolute measures are expressed in the same units as the raw data and describe the actual scatter of values. Relative measures are dimensionless ratios that allow fair comparison across datasets with different units or different means.

Range — Definition, Formula, and Examples

The range is the simplest measure of dispersion: the difference between the maximum and minimum values in a dataset. You can calculate it in seconds, which makes it a useful first look at any data — but it tells you nothing about what happens between the extremes.

For grouped data with classes 10–20, 20–30, 30–40, 40–50: Range = 50 − 10 = 40.

Coefficient of Range (Relative Measure)

Advantages and Disadvantages of Range

Interquartile Range (IQR) and Quartile Deviation

The interquartile range measures the spread of the middle 50% of data — the range between the first quartile (Q1) and the third quartile (Q3). Because it ignores both tails, it resists the influence of outliers and is the natural partner to the median.

How to Find Q1, Q2, and Q3

Quartiles divide ordered data into four equal parts. Their positions are:

• Q1 (25th percentile): position (n+1)/4
• Q2 (50th percentile, median): position (n+1)/2
• Q3 (75th percentile): position 3(n+1)/4

When a position falls between two values, interpolate by averaging the two surrounding values. For a deeper walkthrough of how percentiles relate to quartiles, see percentiles.

Quartile Deviation (Semi-IQR)

Also called the semi-interquartile range, this tells you the typical distance from the median to either quartile. The relative version for cross-dataset comparison:

Five-Number Summary and Box Plots

The five-number summary {Min, Q1, Q2, Q3, Max} is the natural extension of quartiles. In a box plot, the box spans Q1 to Q3 (the IQR), and the whiskers extend to the minimum and maximum non-outlier values. Any point beyond 1.5 × IQR from Q1 or Q3 is typically flagged as a potential outlier.

Advantages and Disadvantages of IQR

See the full interquartile range page for additional examples, including the IQR method for outlier detection.

Mean Deviation — Definition, Formula, and Examples

Mean deviation (also called mean absolute deviation or MAD) is the arithmetic mean of the absolute differences between each data point and a central value — usually the mean or the median. It is more intuitive than variance because it stays in the original units and avoids squared values.

Why Absolute Values?

Without absolute values, positive and negative deviations cancel exactly. For any dataset, the deviations from the mean always sum to zero. The absolute value keeps the sign from destroying the information.

Mean Deviation About the Mean (Ungrouped Data)

Mean Deviation About the Median

Mean deviation is minimized when calculated about the median. This is why some analysts prefer this version — the median minimizes the sum of absolute deviations, making M.D.(Median) the smallest possible value for any central reference point you choose.

Mean Deviation for Grouped Data

MAD vs. Standard Deviation — Why Statistics Prefers SD

Standard deviation is preferred over MAD not because it is more intuitive, but because squared deviations have superior mathematical properties. They are differentiable, additive across independent groups, and they link directly to the variance used in every major statistical test.

Variance — Definition, Formula, and Examples

Variance is the average of the squared deviations from the mean. Squaring solves the sign cancellation problem (negative and positive deviations no longer cancel) while maintaining the algebraic properties that make further statistical work possible. For a standalone treatment, see the variance page.

Population Variance

Sample Variance (Bessel's Correction)

Why n−1 Instead of n? — Bessel's Correction Fully Explained

Variance for Grouped Data

Step-Deviation Method for Grouped Data

When class widths are equal, this shortcut saves significant arithmetic. Define a coded variable uᵢ for each class:

The Problem with Squared Units

Variance is expressed in squared units. If your data is in metres, variance is in metres². A variance of 9 m² does not mean "9 metres of spread" — it means the squared average deviation is 9 m². This is why the standard deviation, which takes the square root and restores the original units, is what gets reported in most practical contexts. Variance remains essential inside formulas for ANOVA, regression, and related methods — see ANOVA.

Properties of Variance

• Non-negative: σ² ≥ 0 always. If σ² = 0, every value in the dataset is identical.
• Adding a constant: Var(X + c) = Var(X). Shifting all values by the same amount does not change variance.
• Multiplying by a constant: Var(cX) = c² × Var(X). Scaling by c multiplies variance by c².
• Independent samples: Var(X + Y) = Var(X) + Var(Y) when X and Y are independent.

When Variance > SD and When SD > Variance

Standard Deviation — Definition, Formula, and Examples

Standard deviation is the positive square root of variance. It restores the measure of dispersion to the original units of the data, making it directly interpretable. It is the most widely used measure of spread for continuous, approximately symmetric data. For a standalone treatment, see the standard deviation page.

Population Standard Deviation

Sample Standard Deviation

Step-by-Step Calculation (6 Steps)

Standard Deviation for Grouped Data

Standard Deviation and the Empirical Rule

For data that follows a normal distribution, the empirical rule tells you exactly how much of the data falls within each band of standard deviations:

Standard Deviation vs. Standard Error — Full Comparison

Confusing these two is one of the most common errors in published research. They answer different questions entirely.

See standard deviation vs. standard error and the standard error page for further detail.

Properties of Standard Deviation

• Always ≥ 0; equals zero only when all values are identical.
• Uses every data point in its calculation.
• Same units as the original data (unlike variance).
• Adding a constant c to every value leaves σ unchanged.
• Multiplying every value by a constant c multiplies σ by |c|.
• It is not an unbiased estimator of σ (unlike s² for σ²), though the bias is negligible for large n.

Advantages and Disadvantages of Standard Deviation

Coefficient of Variation (CV)

The coefficient of variation is the only dispersion measure that is truly dimensionless. By expressing standard deviation as a percentage of the mean, it lets you compare variability across datasets that have different units or very different means — something no absolute measure can do. See the standalone coefficient of variation page for worked examples.

When to Use CV

• Comparing variability of datasets with different units (e.g., heights in cm vs. weights in kg)
• Comparing datasets with very different means (e.g., salaries in two countries)
• Investment analysis: comparing risk per unit of return across assets

When NOT to Use CV

• When the mean is zero or near-zero (CV becomes infinite or meaningless)
• When data is on an interval scale without a true zero (e.g., temperature in Celsius — a 0°C mean does not mean "no temperature")
• When data contains negative values

All Formulas at a Glance

The tables below consolidate every formula covered on this page for quick reference.

Absolute Measures

Relative Measures

Complete Grouped Data Example

This section works through one frequency distribution and calculates Range, Mean Deviation, Variance, and Standard Deviation using both the direct method and the step-deviation method. Each formula produces an estimate because individual values are collapsed into class intervals.

Chebyshev's Theorem — Dispersion Without Assuming Normality

The empirical rule (68–95–99.7) applies only to normal distributions. For data that is skewed, heavy-tailed, or simply unknown in shape, Chebyshev's theorem provides a guaranteed minimum bound that holds for any distribution with a finite mean and variance.

If you know your data is not normally distributed, use Chebyshev's bounds to make conservative probability statements without assuming any specific distribution shape. Try the Chebyshev's Theorem calculator to apply these bounds to your own data.

Which Measure of Dispersion Should You Use?

The right choice depends on three things: the type of data, whether the distribution is symmetric or skewed, and whether outliers are present. This table handles all three simultaneously.

The Key Rule: SD vs. IQR

When All Measures Disagree

When range, IQR, and SD paint different pictures, graph your data first. A large range combined with a small IQR tells you outliers are the story — report IQR as primary. When SD is noticeably larger than IQR/1.35, the data is right-skewed and IQR should lead. When SD ≈ IQR/1.35, the distribution is approximately normal and either measure is valid. See the histogram and skewness and kurtosis pages for visual diagnostic tools.

How Outliers Affect Each Measure

For a systematic approach to identifying and handling outliers, see the outliers page.

Calculating Measures of Dispersion in Python and Excel

Python

import numpy as np
from scipy import stats
import statistics

data = [12, 14, 15, 16, 18, 19, 20, 22]

# Range
data_range = np.max(data) - np.min(data)
print(f"Range: {data_range}")                 # 10

# IQR and Quartile Deviation
q1 = np.percentile(data, 25)
q3 = np.percentile(data, 75)
iqr = q3 - q1
qd  = iqr / 2
print(f"Q1={q1}, Q3={q3}, IQR={iqr}, QD={qd}")

# Mean Absolute Deviation
mean = np.mean(data)
mad  = np.mean(np.abs(data - mean))
print(f"Mean Absolute Deviation: {mad:.4f}")

# Population Variance and SD (divides by n)
pop_var = np.var(data)
pop_sd  = np.std(data)
print(f"Population Variance: {pop_var:.4f}")
print(f"Population SD: {pop_sd:.4f}")

# Sample Variance and SD (Bessel's correction: ddof=1)
samp_var = np.var(data, ddof=1)
samp_sd  = np.std(data, ddof=1)
print(f"Sample Variance: {samp_var:.4f}")
print(f"Sample SD: {samp_sd:.4f}")

# statistics module (always uses n-1)
print(f"statistics.variance: {statistics.variance(data):.4f}")
print(f"statistics.stdev:    {statistics.stdev(data):.4f}")

# Coefficient of Variation
cv = (samp_sd / mean) * 100
print(f"CV: {cv:.2f}%")

# Standard Error of the Mean
sem = samp_sd / np.sqrt(len(data))
print(f"Standard Error: {sem:.4f}")

Excel

For more on statistical computing, see standard deviation vs. variance for a deeper comparison, or use the descriptive statistics calculator for an all-in-one tool.

Central Tendency and Dispersion — You Always Need Both

The natural pairings reflect shared assumptions about scale and robustness:

Measures of Dispersion in Real Life

Common Mistakes and Misconceptions

Frequently Asked Questions

Key Takeaways