What considerations are required when choosing a statistical test for a dataset?

This question gets asked regularly by A Level and University students. There are a few things to consider when choosing a statistical test for data. Note that more complex statistical tests such as ANOVA/MANOVA, Kruskal Wallis and Friedman’s are beyond the scope of any A-Level syllabus.Firstly, determine whether your dataset consists of scores or frequencies. If it consists of frequencies, a Chi Square should be used.Secondly, consider the experimental design. A between-subjects design will require a different statistical test to a within-subjects design. Note that when the design is between-subjects, subjects only encounter one condition of the independent variable. In a within-subjects design, all subjects encounter all conditions of the IV.Next, determine how many conditions are being tested (i.e. 2-sample or k-sample). A k-sample means that there are more than 2 conditions of the IV.Finally, consider whether your dataset meets assumptions for parametric tests. Generally, this can be gauged based on whether the data is normally distributed and nominal/scalic. If this is the case, then a parametric test should apply. Otherwise, choose an alternative test such as a Wilcoxon or Mann-Whitney.