Which Statistical Test Should I Use?

If you want to summarize one numeric variable, start with descriptive statistics.

If you want to compare means, consider a t-test. If you want association between numeric variables, consider correlation or regression. If you want categorical count differences, consider chi-square.

Numeric outcomes support means, standard deviations, t-tests, correlation, and regression when assumptions fit.

Categorical counts support chi-square tests when expected cell counts are large enough.

Independence, sampling method, distribution shape, outliers, and expected counts can matter more than the final p-value.

A statistically significant result can still be practically small or caused by biased data collection.

Measurements from the same person before and after an intervention are paired; measurements from unrelated groups are independent. Treating paired observations as independent discards the within-pair relationship and can change the uncertainty estimate.

Write down the observational unit, grouping variable, outcome, and whether any subject appears more than once before choosing a calculator mode.

A test choice does not determine whether the sample can detect a practically important effect. Define the smallest meaningful difference, plausible variability, and planned sample size before collecting data when possible.

Report effect sizes and confidence intervals with p-values. A small p-value does not make a tiny effect important, and a large p-value does not prove that two groups are equivalent.