Author(s):
1. Mohamed Ali HAFDI:
High school of technology Laayoune & IMI laboratory, Ibn Zohr university,Morocco
Abstract:
Many tests based on alternative models are considered to test the proportionality assumption for specified covariate.
The common point of all these alternatives is the linear form of an added component according to the covariate
concerned by the test while the proportionality can be also violated by a non linear form of this added component.
In this paper, we propose a test for proportional hazards assumption for specified covariates. The test is based on a
new alternative which can be applied in the two situations cited above and also when the hazards rates under different
values of covariates is not only constant as in the Cox model, but it may cross, go away, and may be monotonic with
time. The limit distribution of the test statistic is derived. Finite samples properties of the test power are analyzed by
simulation and compared with those of existing tests. Application on Real data examples are considered.
Page(s):
435-449
Published:
Journal: Pakistan Journal of Statistics and Operation Research, Volume: 17, Issue: 2, Year: 2021
Keywords:
AndersonDarling test
,
Proportionality assumption
,
Partial likelihood
,
Cox model
References:
References are not available for this document.
Citations
Citations are not available for this document.