Statistics

Learn statistics by starting with the question

Statistical tools turn a question into something that can be calculated. Identify the units, comparison and assumptions before choosing a formula.

Web edition · 14 October 2026

Which tool fits which question?

Choosing a starting point
QuestionStarting pointLimit
How far is a value from the mean?Z-scores and the normal distributionStandardising a value does not establish a normal distribution.
How can a difference between two means be examined?t and F tablesThe design and assumptions matter.
How do event rates in two groups compare?Absolute and relative riskCompare the same event over the same period.

Archived tables and present-day use

The ELYADAL statistical tables document is preserved as a historical source. Check which probability a printed table gives: the left tail, right tail and area between the mean and z are different. Looking up the same number in a different table and obtaining another result may not be a calculation error.

Open the archived statistical tables PDF. Our current web explanations do not present this document as a new edition. The explanations and tools have their own dates.

Include the calculation in your report

State the formula, the units of the inputs, the sample size and the distribution you used. A number without an explanation of how it was calculated is not an account the reader can check. These tools are for learning; they do not automatically select a method for your dataset.

Continue reading

Sources and provenance

Source check / web edition: 14 October 2026. Historical documents retain their own original dates.

Keep the same mean and change the spread

Two sets of observations can have the same average while describing different variation. Calculate both location and spread before calling them similar. The following waiting times, in minutes, are invented teaching values, not participant data, a psychological study or a measurement of any real service.

Work through both small datasets

Set A is 2, 4, 6. Set B is 0, 4, 8. Each contains three values. For A, the sum is 12 and 12 ÷ 3 = 4 minutes. For B, 0 + 4 + 8 = 12, so its mean is also 4 minutes. In ascending order the middle value is 4 in both sets, making both medians 4. Equal means and medians do not make the observations identical.

  1. Calculate the range: largest minus smallest. A gives 6 − 2 = 4 minutes; B gives 8 − 0 = 8 minutes. This measure uses only the two extremes, not how every value is arranged.
  2. Find deviations from the mean: A gives −2, 0, 2; B gives −4, 0, 4. Their sums are zero, so simply adding deviations would hide the variation.
  3. Square and add: A gives 4 + 0 + 4 = 8; B gives 16 + 0 + 16 = 32. The squared units are minutes squared.
  4. Use the stated definition: here we calculate sample variance with denominator n − 1. Since n = 3, divide by 2: A has variance 4 min² and B has variance 16 min². Using denominator n would describe a different convention; name it rather than mixing results.
  5. Take the square root: sample standard deviation is 2 minutes for A and 4 minutes for B. It restores the original unit. These are arithmetic results for the invented lists, not estimates of a real population’s behaviour.

Write the conclusion at the right scale

A complete sentence is: “Both invented sets have mean and median 4 minutes; B has the larger range and sample standard deviation.” Do not conclude that B is less fair, that a difference is statistically significant or that either distribution is normal. Three chosen numbers cannot establish those claims or reveal the sampling process.

Reusable record: question ___; unit ___; ordered values ___; n ___; mean ___; median ___; range ___; variance definition and denominator ___; standard deviation ___; visible pattern ___; unknown sampling context ___. Check the arithmetic independently before interpreting it. NIST’s scale reference explains these measures and their differing sensitivity to extremes; the plotted examples here are our own.

A has values 2, 4, 6 and B has 0, 4, 8. Both means are 4; B has twice the range and sample standard deviation.
Invented values: A range 4, sample SD 2; B range 8, sample SD 4 minutes. © ELYADAL · CC BY 4.0.

NIST/SEMATECH e-Handbook · Measures of Scale