Choosing inferential tests

Exam items rarely ask you to compute anything — they hand you a scenario and ask which test fits. Master one decision table and you can answer a whole cluster of research questions on sight.

The choice turns mostly on how many groups/means you are comparing and whether the data are numeric scores or categories/frequencies.

Use a t-test to compare two means — exactly two groups (e.g., treatment vs. control post-test scores). An independent-samples t-test compares two separate groups; a paired/dependent t-test compares the same group twice (pre vs. post).

Use ANOVA (analysis of variance) when you compare three or more means — it protects against the inflated error you'd get from running many t-tests. ANOVA yields an F ratio; a significant F says at least one group differs, so you follow up with post-hoc tests (e.g., Tukey) to find which. Shortcut: 2 means → t; 3+ means → ANOVA.

Use chi-square (χ²) for categorical / frequency (nominal) data — counts in categories, not means (e.g., is diagnosis category associated with dropout yes/no?). If your scenario counts how many fall in each box, think chi-square.

Use regression for prediction — predicting a continuous outcome from one or more predictors (e.g., predicting GPA from study hours). Multiple regression uses several predictors. Keyword 'predict' signals regression.

Trigger words are your friend: compare 2 means → t-test; compare 3+ means → ANOVA; frequencies/categories → chi-square; predict an outcome → regression.

Clinical Pearl: Match the verb: two means = t-test, three-or-more means = ANOVA, category counts = chi-square, 'predict' = regression. ANOVA (not repeated t-tests) is the safeguard when you have 3+ groups.