Structural Models for Quasi-dependent Failure and Truncation Times

• Activity Number: 343
• Type: Contributed
• Date/Time: Wednesday, August 11, 2004 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Emily C. Martin*+ and Rebecca A. Betensky
• Companies: Harvard School of Public Health and Harvard School of Public Health
• Keywords: quasi-independence test ; truncated data ; structural model

Abstract

Randomly truncated survival data arise when the failure time is observed only if it falls within a subject-specific truncating interval. Most estimators of the survival function and regression models based on such data rely on the key assumption of quasi-independence, i.e., factorization of the joint density of failure and truncation times into a product proportional to the individual densities in the observable region, as well as the usual assumption of independent censoring. We propose semiparametric structural models for latent failure or truncation times applicable under quasi-dependence of observed failure and truncation times and independent censoring. Estimation is based on either observed failure time conditional on latent truncation time, or latent failure time conditional on observed truncation time, as determined by the chosen structural model. A quasi-independence test statistic that conditions on covariates of interest provides an estimating function for model parameters. Modifiers of the effect of truncation on failure appear in the structural model for the latent random variable. The approach is illustrated using real datasets and performance is simulated.