, Jianwen Cai and Yu Shen
University of North Carolina at Chapel Hill and
The University of Texas
Abstract: Interval-censored event time data often arise in medical and public health studies. In such a setting, the exact time of the event of interest cannot be observed and is only known to fall between two monitoring times. Our interest focuses on the estimation of the effect of risk factors on interval-censored data under the semiparametric additive hazards model. A nonparametric step-function is used to characterize the baseline hazard function. The covariate coefficients are estimated by maximizing the observed likelihood function, and their variances are obtained using the profile likelihood approach. We show that the proposed estimates are consistent and have asymptotic normal distributions. We also show that the estimator obtained for the covariate coefficient is the most efficient estimator. Simulation studies are conducted to assess the performance of the estimate. The method is illustrated through application to a data set from an HIV study.
Key words and phrases: Additive hazards regression, interval-censored data, nonparametric maximum likelihood estimates, profile likelihood, semiparametric efficiency.