Alternative Forms of the Score Test for Heterogeneity in a Censored Exponential Model

Different versions of the score test for neglected heterogeneity for a right censored exponential model are analyzed. These tests depend on how the information matrix is estimated. A test based on the theoretical information matrix is derived that is shown to outperform all the other tests. Further, the noncentrality parameter of the test is examined to show how the power of the test is reduced when data are censored. © 1997 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology