Sankhya: The Indian Journal of Statistics

2005, Volume 67, Pt. 4, 722--735

A Bartlett Type Correction for Rao's Score Test in Cox Regression Model

Dongsheng Tu, Queen's University, Kingston, Canada
Jiahua Chen, University of Waterloo, Waterloo, Canada
Peide Shi, Peking University, Beijing, P.R. China
Yaohua Wu, University of Science and Technology of China, Hefei, P.R. China

SUMMARY. Gu and Zheng (1993) established a Bartlett adjustment for the partial likelihood ratio test in the Cox proportional hazards regression model.  It is known that the same method of adjustment cannot be used to improve the accuracy of the test statistic derived by Rao's score method.  In this paper, the method of Cordeiro and Ferrari (1991) is used to derive a Bartlett type correction for the score test in the Cox proportional hazards regression model.  As a special case, a test which is more accurate than the popular log-rank test is obtained.

AMS (2000) subject classification. Primary 62E20, 62G05; secondary 62P10.

Key words and phrases. Asymptotic expansions, Bartlett adjustment, proportional hazards models, random censoring model, Rao's score test.

Full paper (PDF)