Abstract: It is common in regression analysis of failure time data, such as the AIDS Clinical Trail Group (ACTG) 175 clinical trial data, that the failure time (AIDS incidence time) is subject to interval-censoring and the covariate (baseline CD4 count) is subject to measurement error. To perform valid analysis in this setting, we propose a functional inference method under the semiparametric proportional odds model. The proposal utilizes the working independence strategy to handle general mixed case interval censorship, as well as the conditional score approach to handle mismeasured covariate without specifying the covariate distribution. The asymptotic theory, together with a stable computational procedure combining the Newton-Raphson and self-consistency algorithms, is established for the proposed estimation method. We illustrate the performance of the proposal via simulation studies and analysis of ACTG 175 data.

Key words and phrases: Conditional score, interval-censoring, measurement error, semiparametric, survival analysis.