Abstract.
We consider a semiparametric competing risk model given by k independent survival times. The paper offers an asymptotic treatment of tests for the semiparametric null hypothesis of equality of the underlying risks. It turns out that modified rank tests are asymptotically efficient for certain semiparametric submodels, where the baseline hazard is a nuisance parameter. In addition, the asymptotic relative efficiency of the present tests is derived. A comparison of asymptotic power functions can then be used to classify various tests proposed earlier in the literature. For instance a chi-square type test is efficient for proportional hazards. Data driven tests of likelihood ratio type are proposed for cones of alternatives. We will consider certain stochastically increasing alternatives as a special example. The paper shows how the concept of local asymptotic normality of Le Cam works for hazard oriented models.
Similar content being viewed by others
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Beyersmann, J., Janssen, A. & Mayer, CD. Efficient rank tests for semiparametric competing risk models. Metrika 60, 73–91 (2004). https://doi.org/10.1007/s001840300297
Issue Date:
DOI: https://doi.org/10.1007/s001840300297