Table 1.

Most important statistical tests

Statistical test	Description
Fisher’s exact test	Suitable for binary data in unpaired samples: The 2 × 2 table is used to compare treatment effects or the frequencies of side effects in two treatment groups
Chi-square test	Similar to Fisher’s exact test (albeit less precise). Can also compare more than two groups or more than two categories of the outcome variable. Preconditions: sample size ca. >60; expected number in each field 5
McNemar test	Preconditions similar to those for Fisher’s exact test, but for paired samples
Student’s t-test	Test for continuous data. Investigates whether the expected values for two groups are the same, assuming that the data are normally distributed. The test can be used for paired or unpaired groups
Analysis of variance	Test preconditions as for the unpaired t-test, for comparison of more than two groups. The methods of analysis of variance are also used to compare more than two paired groups
Wilcoxon’s rank sum test (also known as the unpaired Wilcoxon rank sum test or the Mann–Whitney U test)	Test for ordinal or continuous data. In contrast to Student’s t-test, does not require the data to be normally distributed. This test too can be used for paired or unpaired data
Kruskal–Wallis test	Test preconditions as for the unpaired Wilcoxon rank sum test for comparing more than two groups
Friedman test	Comparison of more than two paired samples, at least ordinal scaled data
Log rank test	Test of survival time analysis to compare two or more independent groups
Pearson correlation test	Tests whether two continuous normally distributed variables exhibit linear correlation
Spearman correlation test	Tests whether there is a monotonous relationship between two continuous, or at least ordinal, variables