Analysis of a fixed center effect additive rates model for recurrent event data

Abstract

A center effect additive rates model is suggested to analyze recurrent event data. The proposed model is a useful alternative to the center effect proportional rates model and provides a direct interpretation of parameters. The traditional estimation methods treat the centers as categorical variables, and they comprise many parameters when the number of centers is large and thus may not be feasible in many situations. An estimation method based on the difference in the observed to the expected number of recurrent events is recommended to address the deficiency of the traditional method. The asymptotic properties of the proposed estimator are established. We develop a goodness-of-fit test for model checking. Simulations are conducted to evaluate the small sample performance and show the computational advantage of the suggested method. The proposed methodology is applied to the Childhood Cancer Survivor Study.