power: a reproducible research tool to ease monte carlo power simulation studies for goodness-of-fit tests in r

Abstract

The PoweR package aims to help obtain or verify empirical power studies for goodnessof-fit tests for independent and identically distributed data. The current version of our package is only valid for simple null hypotheses or for pivotal test statistics for which the set of critical values does not depend on a particular choice of a null distribution (and on nuisance parameters) under the non-simple null case. We also assume that the distribution of the test statistic is continuous. As a reproducible research computational tool it can be viewed as helping to simply reproducing (or detecting errors in) simulation results already published in the literature. Using our package helps also in designing new simulation studies. The empirical levels and powers for many statistical test statistics under a wide variety of alternative distributions can be obtained quickly and accurately using a C/C++ and R environment. The parallel package can be used to parallelize computations when a multicore processor is available. The results can be displayed using LATEX tables or specialized graphs, which can be directly incorporated into a report. This article gives an overview of the main design aims and principles of our package, as well as strategies for adaptation and extension. Hands-on illustrations are presented to help new users in getting started.