How many groups or variables are you comparing? Learn the golden rule to choose between Student's t, Mann-Whitney, ANOVA, or Kruskal-Wallis.
¿Cuántos grupos o variables estoy mirando?
Define si usas una prueba simple o compleja al comparar muestras en tu investigación.
t de Student (si es numérica paramétrica) o U de Mann-Whitney (si no es paramétrica).
ANOVA (si es paramétrica) o su equivalente Kruskal-Wallis (no paramétrica).
One of the most frequent challenges for undergraduate and graduate researchers in their methodological framework is selecting the correct statistical hypothesis test. Thesis evaluation committees often reject proposals because a two-sample test was erroneously applied to three groups, or parametric tests were used without validating normality.
The Golden Question: How Many Groups Are You Analyzing?
Once your research objective is clearly set (comparing, correlating, or predicting), you must ask: How many groups or time points am I analyzing?
- If you compare exactly two groups: You need a simple test. If your dependent variable is continuous and meets normality assumptions, use Student's t-test. If data is ordinal or non-normal, use the Mann-Whitney U test.
- If you compare more than two groups (three or more): You face a complex comparison. Do not run multiple t-tests due to family-wise error inflation. You must apply a one-way ANOVA, or its non-parametric counterpart, the Kruskal-Wallis test.
