|
Abstract:
|
We consider temporal process regression of a binary response, for use in studying prevalence (e.g., the prevalence of cancer not-in-remission). Semiparametric model assumptions are built on the expectation of this response, with the expectation representing the probability of being prevalent as a function of covariates. The assumed model is characterized by multiplicative covariate effects and a baseline relative risk function that is unspecified. We show that the regression estimator is asymptotically normal, and the baseline prevalence estimator converges weakly to a Gaussian process. Simulations reveal that the proposed methods have satisfactory finite sample performance. We apply the proposed methods to Scientific Registry of Transplant Recipients (SRTR) data.
|
