Comparing Statistical Software Packages: The Case of the Logrank Test in StatXact

Abstract

This article can be read as a companion to an earlier article by R. A. Oster in which StatXact, LogXact, and Stata were examined. In most cases, comparing software packages on their main features, such as accuracy, user friendliness, and adequate documentation, is sufficient for giving the reader interesting and useful information. In some instances, however, one has to take a different perspective for discovering the mechanism behind discrepancies. An excellent software package, using accurate computations and ample documentation, may nevertheless not do what its menu implies. The case of the logrank test in StatXact is a good example, illustrating the need to always check procedures and formulas against labels and names.