Descriptive Statistics Measures — Complete Comparison Guide

By CalcMulti Editorial Team··9 min read

Descriptive statistics summarise a dataset using two types of measures: central tendency (where is the centre of the data?) and spread (how dispersed is the data?). Choosing the right measure depends on your data type, distribution shape, and what question you are asking.

The most common mistake: using the mean and standard deviation for skewed data with outliers, when median and IQR would give a much more honest picture. This guide compares all key descriptive measures and tells you exactly when each is appropriate.

Central Tendency Measures
VS
Spread Measures

Side-by-Side Comparison

MeasureTypeFormula/MethodSensitive to Outliers?Best For
MeanCentral tendencyΣx / nYes — stronglySymmetric, approximately normal data
MedianCentral tendencyMiddle value when sortedNo — robustSkewed data, income, prices, ordinal
ModeCentral tendencyMost frequent value(s)NoCategorical data, multimodal distributions
Standard DeviationSpread√[Σ(x−x̄)²/(n−1)]YesSymmetric, normal data alongside mean
IQRSpreadQ3 − Q1No — robustSkewed data or outliers alongside median
RangeSpreadmax − minExtremely sensitiveQuick overview; avoid for skewed data
Coefficient of VariationSpread (relative)(SD / mean) × 100%YesComparing spread across different scales
VarianceSpreadSD²YesMathematical analysis (ANOVA, regression)
PercentilesPosition/spreadValue at given rankNoGrowth charts, standardised tests, distributions

Central Tendency — Mean vs Median vs Mode

The mean is the arithmetic average: sum of all values divided by n. It is the most commonly reported measure and has the best mathematical properties — it minimises the sum of squared deviations. But it is sensitive to extreme values: a single very large or very small value shifts the mean significantly.

The median is the middle value when data is sorted. For odd n, it is the exact middle; for even n, it is the average of the two middle values. The median is robust — extreme values do not affect it at all (as long as they stay on the same side). It is the preferred measure when data is skewed (e.g., income, house prices, reaction times).

The mode is the most frequently occurring value. For continuous data, it is often not useful (each value may appear only once). For categorical data (e.g., most popular product size, most common blood type), the mode is the only sensible central tendency measure.

ScenarioUseReason
Exam scores in a class (normal-ish)MeanSymmetric; mean is most efficient
Household income in a cityMedianRight-skewed; a few billionaires inflate the mean
House prices in a neighbourhoodMedianRight-skewed; outliers distort mean
Most popular T-shirt sizeModeCategorical — mean/median are meaningless for sizes
Age of employees (bimodal)Both mean and modeReport both; describe the two peaks
Response time on a website (skewed)Median + percentilesSkewed; p50, p95, p99 are most informative

Spread Measures — SD, IQR, Range, and More

Standard deviation (SD) is the most commonly used spread measure for approximately normal data. It is in the same units as the data, directly interpretable, and the natural companion to the mean. SD is sensitive to outliers because it depends on the mean.

Interquartile Range (IQR = Q3 − Q1) covers the middle 50% of data. It is completely robust to outliers — extreme values outside the "whiskers" do not change Q1 or Q3. IQR is the natural companion to the median. Used in box plots.

Range (max − min) is the simplest spread measure but the least useful: one extreme outlier can make range enormous. Use it only as a rough orientation when the dataset is small and well-behaved.

Coefficient of Variation (CV = SD/mean × 100%) allows comparing spread across datasets with different units or different means. "Investment A has CV = 15%; Investment B has CV = 40%" — Investment B is more volatile relative to its mean return.

How Outliers Affect Each Measure — Example

Dataset: Monthly salaries of 7 employees (£000s): 30, 32, 33, 34, 35, 36, 200.

Mean: (30+32+33+34+35+36+200)/7 = 400/7 = £57,143. The CEO (£200k) inflates the mean to £57k — not representative of the typical employee.

Median: Sorted: 30, 32, 33, 34, 35, 36, 200. Middle value = 34 = £34,000. Not affected by the outlier at all. Much more representative of a typical employee.

Mode: No mode (all values appear once).

SD: Very large due to the outlier (≈ 58.7k). Mean ± 1 SD = £57k ± £59k — a range of [−£2k, £116k]. Meaningless.

IQR: Q1 = 32, Q3 = 35.5, IQR = 3.5 (£3,500). Not affected by the CEO's salary. Accurately describes that the middle 50% of regular employees earn within a £3,500 range.

Conclusion: For this skewed dataset with an outlier, report median (£34k) and IQR (£3.5k). The mean and SD give a misleading picture.

Matching Measures to Distribution Shape

The distribution shape is the most important factor in choosing which measures to report.

Distribution ShapeCentral TendencySpreadVisualisation
Symmetric, normalMeanStandard DeviationHistogram, bell curve
Right-skewed (long right tail)MedianIQRBox plot, histogram
Left-skewed (long left tail)MedianIQRBox plot, histogram
Bimodal (two peaks)Both modes + meanSD of each peak if separableHistogram showing both peaks
Heavily tailedMedianIQR or percentilesBox plot with outliers shown
Categorical (nominal)ModeFrequency tableBar chart
OrdinalMedianIQR, percentilesOrdered bar chart, box plot

The Five-Number Summary — A Complete Picture

The five-number summary provides a complete descriptive picture of a dataset without assuming any particular distribution: Minimum, Q1, Median (Q2), Q3, Maximum.

These five values are the basis of the box plot (box-and-whisker plot). Together they show: the range (min to max), the centre (median), the spread of the middle 50% (box from Q1 to Q3), and the presence of outliers (points beyond the whiskers at Q1 − 1.5×IQR and Q3 + 1.5×IQR).

When to use the five-number summary vs mean ± SD: always use the five-number summary when data is skewed, when outliers are present, or when you want to show the full distribution without assuming normality. Use mean ± SD for symmetric, approximately normal data where the empirical rule (68%/95%/99.7%) is meaningful.

Summary

The right descriptive statistics depend on your data's distribution shape and type. For symmetric normal data: report mean and SD. For skewed data or data with outliers: report median and IQR. For categorical data: report mode and frequencies.

  • Mean and SD: use for symmetric, approximately normal, continuous data
  • Median and IQR: use for skewed distributions, data with outliers, or ordinal data
  • Mode: use for categorical data or to describe the most common value in multimodal distributions
  • Always plot your data before choosing measures — a histogram reveals skewness and outliers
  • Never report mean and SD for clearly skewed data — it gives a misleading picture of the typical value

Frequently Asked Questions

Educational use only. Content is based on publicly documented mathematical formulas and reviewed for accuracy by the CalcMulti Editorial Team. Last updated: February 2026.