How to choose the right statistical test

The most common mistake in quantitative analysis isn't a wrong calculation, but a wrong test choice. The right test depends on three things: your variables, your hypotheses and your data type.

1. Identify the variable type

• Nominal categorical: no order (e.g., gender, sector).
• Ordinal categorical: with order (e.g., education level).
• Continuous numeric: (e.g., income, time).

2. Define your goal

Compare groups, relate variables or predict? Each goal opens a family of tests:

• Compare 2 groups (numeric): Student's t or Mann-Whitney U.
• Compare 3+ groups (numeric): ANOVA or Kruskal-Wallis.
• Associate 2 categoricals: Chi-square.
• Relate numerics: Pearson or Spearman correlation.
• Predict: linear/logistic regression or ML models.

3. Check the assumptions

Parametric tests (t, ANOVA, Pearson) require normality and homoscedasticity. If they don't hold, use the non-parametric equivalent. Skipping this step invalidates the result.

4. Report with sense

The p-value isn't enough: report the statistic, degrees of freedom, confidence interval and effect size. A careful reviewer will ask for them.