How to choose your study sample and calculate its size
Probability and non-probability sample types, and Cochran's formula for calculating the size, with an interactive calculator that computes your sample instantly.
Choosing the wrong test is one of the things defense committees criticize most. The rule: the test is determined by the type of hypothesis (difference, relationship or effect), the type of variables (quantitative or categorical), the number of groups, and whether the assumptions are met.
| What you want to test | Variables | Parametric test | Non-parametric alternative |
|---|---|---|---|
| A difference between two independent groups | Categorical + quantitative | Independent t-test | Mann-Whitney |
| A difference in the same group before and after | Two quantitative measures | Paired t-test | Wilcoxon |
| A difference between three or more groups | Categorical + quantitative | One-way ANOVA | Kruskal-Wallis |
| A relationship between two quantitative variables | Quantitative + quantitative | Pearson | Spearman |
| A relationship between two categorical variables | Categorical + categorical | Chi-square | — |
| The effect of several independent variables | Quantitative | Multiple regression | — |
| A model with multiple and mediating relationships | Latent | SEM (AMOS / SmartPLS) | PLS-SEM |
Check the assumptions first
- Normality: the Kolmogorov-Smirnov or Shapiro-Wilk test, or skewness and kurtosis values.
- Homogeneity of variance: Levene's test before a t-test and ANOVA.
- In regression: no multicollinearity (VIF below 10, preferably below 5), and independence of residuals.
- If the assumptions aren't met, use the non-parametric alternative or a method that doesn't require normality.
Was this guide useful? Share it with your colleagues.