Non-parametric Within-Subjects Test Parametric assumptions · Normal distribution · Homogeneity of variance · Independence of data points If data violates parametric assumptions . Check for and deal with outliers · Transform the data . Run an alternative test If you run a parametric test when data violates assumptions · Compromises the precision · Inaccurate behavioural predictions Statistical power . The probability of finding a significant effect · Typically, researchers ai for 80% power For a non-parametric within-subjects test you will use Wilcoxon Matched Pairs Signed Rank Test / Wilcoxon When to use Wilcoxon · When testing differences between two conditions · With data that does not meet parametric assumptions · Within-subjects design Wilcoxon ranks the differences across conditions from smallest to largest You could also use a Friedman test When to use a Friedman test . When testing more than 2 conditions · With data that does not meet parametric assumptions · Within-subjects design Effect size You must report effect sizes if you find a significant effect using the formula r = z / VN Non-significant results Any observed difference is due to chance There either may be a difference that we haven't yet found or there many not be any difference
Here we have failed to reject the null hypothesis · Reject the null hypothesis (p< . 05) - significant · Fail to reject the null hypothesis (p> .05) - not significant Bayesian Methods This is an alternative statistical approach It is possible to evaluate strength of evidence in favour of the null/experimental hypothesis There is current movement in the field towards using this approach