Identifying a Time-dependent Covariate Effect in the Additive Risk Model

by Cyrus Amir and Ian W. McKeague
 
Abstract
 
We develop a method to identify
a  time dependent covariate effect in the partly parametric
additive risk model. The proposed method is based on a formal hypothesis
test, whereas previously only an ad hoc procedure was available.   
Rates of convergence of restricted  maximum likelihood estimators of
regression coefficients based on the method of sieves play an
important role in the development.   We carry out a simulation study
to assess the performance of the proposed test. We also apply our test
to real data from a clinical trial on myelomatosis.