Examqo › Examqo library › Public health › Biostatistics and screening · IMLE
Choosing a statistical test
A test is chosen by three questions: what type of outcome variable (quantitative or categorical), how many groups are compared, and whether the measurements are independent or paired – and if quantitative data are not normal, switch to a non-parametric test.
Why it matters in the exam
Items describe a study – for example comparing mean blood pressure between two treatment groups, comparing the death rate between three hospitals, or a measurement before and after an intervention in the same people – and ask which test fits. The key distinctions: the t-test for two independent groups, the paired t-test for the same people before and after, analysis of variance (ANOVA) for more than two groups, chi-square for comparing proportions, Fisher's exact test when expected frequencies are small, and McNemar for paired categorical data. Every parametric test has a non-parametric counterpart: Mann–Whitney, Wilcoxon and Kruskal–Wallis. Further items ask about the log-rank test for comparing survival curves and about logistic or Cox regression when adjusting for several variables. In these items read the study description carefully: the word "same" hints at a paired measurement, and the word "proportion" hints at a categorical variable. A simple way to cope is to picture a small table: rows for the outcome type, columns for two groups, more than two, or a paired measurement. Almost every item falls into one of the cells, and the trap is usually a test that suits another data type, such as a t-test for proportions or chi-square for means.
Key points
- The outcome type, the number of groups and pairing decide the test.
- The t-test for two groups, ANOVA for more than two.
- Chi-square for proportions, Fisher for small expected frequencies, McNemar for paired categorical data.
- Every parametric test has a non-parametric alternative.
- The log-rank test compares survival curves.
Summary table
| Situation | Normal or categorical data | Non-parametric alternative | Example |
|---|---|---|---|
| Two independent groups, quantitative outcome | Two-sample t-test | Mann–Whitney | Blood pressure on drug versus placebo |
| The same people twice, quantitative outcome | Paired t-test | Wilcoxon signed-rank | Weight before and after a programme |
| Three or more groups, quantitative outcome | Analysis of variance (ANOVA) | Kruskal–Wallis | Haemoglobin on three diets |
| Two or more groups, categorical outcome | Chi-square | Fisher's exact test when expected frequencies are below 5 | Infection rate in two hospitals |
| Paired categorical outcome | McNemar test | – | Results of two tests in the same subjects |
| Time to event | Log-rank test, Cox regression | – | Survival on two treatments |
In the full version for subscribers
- A simple algorithm
- Memory aid
- Exam traps
- 5 review cards
- Enough for IMLE
The full version is open to subscribers of: Public health, Medical licensing exam (IMLE).
Subscriber? Read the full topic Plans and prices 25 free questions
Sources
- Kirkwood & Sterne, Essential Medical Statistics, 2nd ed., 2003 — Comparing means; Comparing proportions; Methods based on ranks
- Altman, Practical Statistics for Medical Research, 1991 — Comparing groups – continuous and categorical data
Sources are given for further reading; the text is written afresh and not copied from them.
Reviewed by the Examqo AI Professor · 2026-11-06. The text may contain errors; when in doubt, check the source.