B0338
Title: Iterative generalized least squares estimation for the analysis of multilevel interval-censored survival data
Authors: Samuel Manda - University of Pretoria (South Africa) [presenting]
Abstract: Many medical areas including dentistry and HIV/AIDS research domains collect time-to-event data that are often interval-censored. In regression modelling of interval-censored failure time data, the Cox proportional hazards model is commonly used. Analyses have been extended to account for the lack of independence in the data, which, for example, may arise from the clustering of observations in multilevel data structures. Estimation is carried out with one of the available methods including marginal likelihood, generalised estimation equations (GEE), the EM algorithm, Bayesian methods and the maximum likelihood of a copula model for multivariate interval survival data. An alternative estimation method is presented based on iterative generalised least squares (IGLS). Two example data sets are used to illustrate the proposed estimation technique.