NonpaR: Nonparametric Tests

Test Selection


Test Selection

The NonpaR module provides several test options. The sections below briefly describe each test and the experimental design that is covered by the test. The NonpaR manual section will provide additional information and a reference to books that describe these methods in greater detail.

Wilcoxon Rank Sum Test -- Two Experimental Groups

The Wilcoxon Rank Sum Test is used to evaluate differences between exactly two experimental groups. The Wilcoxon statistic is linearly related to the U statistic which is produced by the Mann-Whitney test which means that these tests are generally considered equivalent. The null hypothesis is that the population means from which the two samples are draw are the same. The test presents results to help justify rejection of this null hypothesis when appropriate.

Kruskal-Wallace Test -- N Experimental Groups

The Kruskal-Wallace Test is used to look for differences between N experimental groups where N is usually greater than or equal to 3 (for 2 groups the Wilcoxon tests is usually performed). The null hypothesis is that all population means from the populations being sampled are the same. The test hypothesis is that at least one population mean is different from the others.

Mack-Skillings Test -- Two Way Designs: Two Experimental Factors

The Mack-Skillings Test is a generalization of the Friedman tests for two-factor designs. The test covers the case where there are two experimental factors being studied and each factor has multiple levels. An example might be where one factor is a pair of bacterial strains and a second factor is antibiotic concentration of which there are 3 concentrations. This is a 2x3 design, two strains, three antibiotic concentrations. The design can be described in a 2x3 matrix. Unlike the Friedman test, the Mack-Skillings test allows for replication with each cell of the design matrix. This tests will report significant findings with respect to the two factors under study.

Fisher Exact Test -- Two Experimental Groups and Binned Categorical Data

The Fisher Exact Test focuses on analyzing data where there are exactly two experimental groups and the underlying data can be binned into two categories. One example is for CGH were values above or below a given threshold indicate the presence or absence of a gene and the samples can be split into group A and group B. In this example the Fisher Exact test focuses on finding genes that are disproportionately present or absent in group A relative to group B. This test requires that you provide a cutoff value to bin the data values into two groups. A 2x2 matrix records the number of data values for a gene that fall into the two sample groups and the two data bins. p-values from the test describe whether non-random associations exist between these two factors.