THREE TESTING PROCEDURES FOR A COMPETING RISKS MODEL

This paper is concerned with testing for the equality of failure rates in a competing risks model with two risks. Three testing procedures are investigated, namely Score test, Likelihood Ratio test and Wald test. Wald test has been considered to be the most powerful in multivariate linear regression analysis.1 However in our application, Wald test is the most efficient one when the failure rate of the second failure type is strictly smaller than the first failure type, otherwise the Score or the Likelihood Ratio test is preferred. This phenomenon is illustrated by data from a mechanical-switch life test.2 The results has been extended to a k-competing risks model. A simulation study is also given to examine the performance of the three tests.