Compare 3 or more groups

Do the centers of two groups differ?
Parametric Test: Two-sample t-test. Tests differences in means. Requires that both groups follow a normal distribution.
Permutation Test: Two-sample perm test. Tests differences in means. Does not require that both groups follow a normal distribution.
Nonparametric Test. Rank-sum (also called Mann-Whitney) test. Tests differences in medians. No distributional assumption required.

See Helsel & others (2017) Chapter 5.


Do paired data differ?
Parametric Approach: Paired t-test. Tests whether the mean difference is zero. Requires that differences follow a normal distribution.
Permutation Test: Paired perm test. Tests differences in means. Does not require that differences follow a normal distribution.
Nonparametric Approach. Sign or signed-rank test. Tests whether the median difference is zero. No assumption of a normal distribution required.

See Helsel & others (2017) Chapter 6.


Do the centers of three or more groups differ?
Parametric Approach: Analysis of variance. Tests differences between means. Requires that each group's data follow a normal distribution.
Permutation Test: One-factor or multi-factor perm tests. Tests differences in means. Does not require that each group's data follow a normal distribution.
Nonparametric Approach. Kruskal-Wallis test. Tests differences between medians. Does not require that each group's data follow a normal distribution.

See Helsel & others (2017) Chapter 7.

Online at:


Chapter references are to the 2017 Second Edition of Statistical Methods in Water Resources.