Multiple-samples tests

Author

Cox Lab

Published

March 18, 2024

1 General

  • Type: - Matrix Processing
  • Heading: - Tests
  • Source code: not public.

2 Brief description

Multi-sample test for determining if any of the means of several groups is significantly different from each other.

Output: A numerical columns is added containing the p-value. In addition there is a categorical column added in which it is indicated by a ‘+’ when the row is significant with respect to the specified criteria.

3 Parameters

3.1 Grouping

Selected categorical row that defines the grouping of columns that should be used in the test (default: first categorical row in the matrix).

3.2 Test

Defines what kind of test should be applied (default: ANOVA). The test can be selected from a predefined list:

  • ANOVA
  • Kruskal Wallis

3.2.1 S0

Artificial within groups variance (default: 0). It controls the relative importance of t-test p-value and difference between means. At \(s0=0\) only the p-value matters, while at nonzero s0 also the difference of means plays a role. See (Tusher, Tibshirani, and Chu 2001) for details.

3.3 Use for truncation

Defines on what value the truncation is based on (default: Permutation-based FDR). Choose here whether the truncation should be based on the p-values, on permutation-based FDR-values or, if the Benjamini-Hochberg correction for multiple hypothesis testing should be applied.

3.3.1 Threshold p-value

This parameter is just relevant, if the parameter “Use for truncation” is set to “P-value”. Rows with a test result below this value are reported as significant (default: 0.05).

3.3.2 FDR

This parameter is just relevant, if the parameter “Use for truncation” is set to “Benjamini-Hochberg FDR” or “Permutation-based FDR”. Rows with a test result below this value are reported as significant (default: 0.05).

3.3.3 Number of randomizations

Specifies the number of randomizations that should be applied (default: 250).

3.3.4 Preserve grouping in randomizations

Defines, whether the grouping specified in a categorical row should be preserved in the randomizations (default: <None>). It can be selected from a list including all available groupings of the matrix.

3.4 Log10

If checked, \(-Log_{10}(test\ value)\) is reported in the output matrix (default). Otherwise the test-value is reported.

3.5 Suffix

The entered suffix will be attached to newly generated columns (default: empty). That way columns from multiple runs of the test can be distinguished more easily.

4 Parameter window

References

Tusher, Virginia Goss, Robert Tibshirani, and Gilbert Chu. 2001. “Significance Analysis of Microarrays Applied to the Ionizing Radiation Response.” Proceedings of the National Academy of Sciences 98 (9): 5116–21. https://doi.org/10.1073/pnas.091062498.