testforDEP: An R Package for distribution-free tests and visualization tools for independence

This article introduces testforDEP, a portmanteau R package containing several tests and visualization tools to examine independence between two variables. This new package combines classical tests including Pearson’s product moment correlation coefficient method, Kendall’s τ rank correlation coefficient method and Spearman’s ρ rank correlation coefficient method with modern tests consisting of density-based empirical likelihood ratio test, Kallenberg datadriven test, Maximal information coefficient test, Hoeffding’s independence test, empirical likelihood based test, and continuous analysis of variance test. For two variables the function testforDEP provides an interface to those tests and returns test statistics, corresponding p values, and bootstrap confidence intervals. The function AUK provides an interface for Kendall plots and computes the area under the Kendall curve. In this paper, we present the testforDEP package and perform a power analysis via Monte-Carlo simulations ultimately concluding that classical tests are superior for simple linear dependence structures while the more modern tests are more powerful for non-linear and random-types of dependence.