Global tests in the additive hazards regression model

Abstract

In this article, we discuss testing for the effect of several covariates in the additive hazards regression model. Bhattacharyya and Klein (Statist. Med. 2005; 24(14):2235–2240) note that an ad hoc weight function suggested by Aalen (Statist. Med. 1989; 8:907–925) is inconsistent when used as a global test for comparing groups since the test statistic depends on which group is used as the baseline group. We will suggest a simple alternative test that does not exhibit this problem. This test is a natural extension of the logrank test. We shall also discuss an alternative covariance estimator. The tests are applied to a data set and a simulation study is performed. Copyright © 2007 John Wiley & Sons, Ltd.