A New Approach to Estimating the Distribution of HIV Infection Time with Interval Censored Data

Supervisor: X. Joan Hu

The available information on time to HIV infection in an AIDS study is often framed into interval censored data. Well-established approaches such as the Turnbull estimator are believed rather inefficient especially when the censoring intervals are relatively wide. Motivated partly by a wildfire control study, Xiong (PhD thesis, 2020) presents a methodology for making inference on the distribution of an event duration from data with missing time origin. She links available longitudinal measures to the event duration and estimate the duration distribution via the first-hitting-time model (e.g. Lee and Whitmore, 2006). This project aims to adapt Xiong's approaches in AIDS studies where the time to HIV infection is missing and there are available biomarkers over time. It includes the following two stages.

Stage 1. We will explore two to three recent AIDS studies to identify biomarker measures relevant to HIV infection.

Stage 2. We will then model HIV infection time jointly with the identified biomarker measures, and conduct inference on the distribution of the infection time.