Choosing the Right Statistical Test: A Quick Reference Guide

Parametric vs Non‑Parametric Tests

• Parametric tests focus on differences in means.
• Non‑parametric tests focus on differences in medians.

Comparing Medians

• Two groups: use the Mann‑Whitney U test.
• Three or more groups: use the Kruskal‑Wallis test.

Comparing Proportions or Percentages

• Standard approach: chi‑square test of independence.
• Small expected counts (any cell < 5): switch to Fisher's exact test, which provides an exact p‑value for sparse data.

Ordinal Variables (Ordered Categories)

• Examples: education level (no formal, primary, secondary, tertiary) or Likert‑scale responses.
• Use Spearman's rho to assess the relationship or differences between ordinal variables.

Continuous Variables (Interval/Ratio Data)

• Examples: age, height, weight.
• Use Pearson's correlation coefficient to examine linear relationships or trends between two continuous measures.

Practical Tips

• Always check the nature of your data (scale, distribution, sample size) before selecting a test.
• For small sample sizes in chi‑square tables, Fisher's exact test prevents inflated Type I error.
• When dealing with ordered categories, treat them as ordinal rather than nominal to preserve the inherent ranking.

For deeper explanations and worked examples, refer to the full course material mentioned at the end of the transcript.

Select the statistical test that matches your data type—means vs medians, proportions vs counts, ordinal vs continuous—to obtain valid, reliable results.