permutatietoetsen

Permutation tests are statistical hypothesis tests that evaluate the null hypothesis by comparing the observed data to a distribution of a test statistic generated under the null hypothesis. They rely on random permutations of the data to create this reference distribution, making them non - parametric and distribution - free. The core procedure involves shuffling the data labels (or observations) while keeping the observed data values fixed, then calculating the test statistic for each permutation. This creates a permutation distribution of the statistic. The proportion of permutations where the statistic is as extreme or more extreme than the observed statistic under the null hypothesis defines the p - value.

Permutation tests are commonly used when traditional parametric assumptions (such as normality) are not met or when sample sizes are small. They are applicable in scenarios like comparing two groups (either paired or independent samples), testing for an association, or assessing the significance of a model's fit. One advantage is that they are robust to outliers and do not require strong distributional assumptions. However, they can be computationally intensive for large datasets, though modern computing power has mitigated this limitation. Overall, permutation tests offer a flexible and powerful alternative to parametric tests, particularly in exploratory or small - sample analyses.