Non-Parametric Paired Data

Non-parametric paired analyses are approached through the same logic as the parametric paired t-tests: a 2 sample paired analysis can be reduced to a 1 sample test by creating a single distribution of the differences between each pair. If the distribution of the differences are nonnormal, and cannot be normalized by transforming the data to some other ratio scale, a 1 sample non-parametric test would be appropriate. Remember that with non-parametric 1 sample tests we have two choices; the 1 sample sign test or the Wilcoxon sign rank test, each with their appropriate assumptions. If you are uncomfortable with the symmetry assumption of the Wilcoxon method, use the sign test.