In this paper, we develop a weighted permutation (WP) method to construct confidence intervals for regression parameters in relative risk regression models. The WP method is a generalized permutation approach. It constructs a resampled history which mimics the observed history for individuals under study. Inference procedures are based on studentized score statistics that are insensitive to the forms of the relative risk function. This makes the WP method appealing in the general framework of the relative risk regression model. First order accuracy of the WP method is established using the counting process approach with a partial likelihood filtration. A simulation study indicates that the method typically improves accuracy over asymptotic confidence intervals.
Jiang, Wenyu and Kalbfleisch, Jack, "Permutation Methods in Relative Risk Regression Models" (July 2006). The University of Michigan Department of Biostatistics Working Paper Series. Working Paper 60.