Case study: Statistical testing using Beanshell scripting

Keywords:

Mann-Whitney, Kolmogorov Smirnov, Fisher's exact test, Chiq-Square test, scripting, Java API, direct call

Questions 1: Are two numeric sets differentially distributed?

Examples are in perparation

Question 2: Is one or more categories over- or underrepresented?

The following two PROMPT Beanshell scripts are testing wether the differentially expressed (DE) genes are significantly enriched or depleted in the GO category/categories of interest. As input the continceny table is used directly as introduced at the overview page is used.

The scripting files can be executed as described at the help page.

All script files were generated by the PROMPT GUI and copied from the Beanshell macro logger logger box. If you need different tests or custom parameters you can simply run your test once in the GUI and see how PROMPT denotes the syntax of your specific parameter settings.

Contingency tables used: contingency_table1.txt and contingency_table2.txt

prompt_script.bsh
Runs Fisher's exact test. If you get a FEXACT Error 40 then the default R workspace is too small. Solution: Add workspace=8e6 or a greater number to the parameter string of the test i.e. test0.init(result0, "testm,hybrid=FALSE,conf.level=0.95,workspace=8e6",... If you get the FEXACT Error 501, then it is recommended to use the ChiSquare test or another alternative test as the Fisher's exact test is not appropriate for a contingency table with such large numbers.

chisquare_script.bsh
Runs a Chi-Square test.