STATISTICS GUIDE

How to choose a statistical test

The right statistical method starts with the study question, how data were collected, and whether assumptions are reasonable—not just the smallest p value.

FORMULA

t = (estimate − null value) / standard error

WORKED EXAMPLE

Use a one-sample t-test for one sample versus a reference mean, Welch's t-test for two independent means, correlation for paired numeric measurements, and one-way ANOVA for three or more group means.

STEP BY STEP

Describe the data first

Use descriptive statistics to check sample size, center, spread, units, and unusual values before choosing an inference test.

Match method to question

A mean comparison, relationship between pairs, and comparison across several groups are different questions with different models.

Check assumptions and interpretation

P values and confidence intervals depend on independence, model fit, and data quality. Treat a calculator output as analysis support, not a study conclusion.