Applied Medical Informatics (Sep 2011)
Coin’s Package Permutation Tests Graphic User Interface Extension for R Commander
Abstract
Background: For small, skewed, sparse, unbalanced data, or data with lots of ties the exact P values computed with permutation tests can help making valid statistical inferences. In the medical field, especially in exploratory studies, like phase I, phase II trials or correlation studies these techniques can prove of a great help. R is a free software environment for statistical computing and graphicsthat offers this kind of tests through the coin package. R Commander graphical user interface to R is one of the most comprehensive in respect of functions offered. The aim of this work was to create an extension for R Commander that offers an interface for permutation tests, a handful of non parametric tests in a permutation framework, and new options for some of the existing non parametric tests, making access to the coin package functions. Materials and Methods: The extension was compiled with R version 2.10. and tested on with R version 2.12.1, and R Commander 1.6-2. Results: The extension has been released successfully, named RcmdrPlugin.coin, and it is available online on CRAN servers for any R Commander user to download and use. The extension creates a new menu, Coin that provides interface to the following tests: Cochran-Mantel-Haenzsel test, Marginal Homogeneity test, linear by linear association test, chi square test; median test, two sample Wilcoxon test, normal quartile test, two/k sample permutation test, Kruskal-Wallis test, Ansari-Bradley test, Fligner-Killeen test, Spearman test, Maximally Selected Statistics, two/k sample test for censored data, paired-samples Wilcoxon test, and Friedman rank-sum test.
