Nonparametric Method of Mixture Model with Prevalent Sampling

• Activity Number: 184
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Yu-Jen Cheng*+ and Mei-Cheng Wang
• Companies: National Tsing-Hua University and The Johns Hopkins University
• Keywords: Left truncation ; mixture model ; nonparametric MLE

Abstract

The objective of this article is to make inference on the survival function and the cure probability subject to left truncation. The problem is especially complex because death and cure are contrast events and both events could be truncated before data recruitment. Mixture model is considered in this article. We addressed the connection between mixture model and competing risk model under prevalent sampling scheme and developed a nonparametric approach to estimate the survival function based on a weaker assumption, conditional independence. Nonparametric MLE of the survival function and the probability of cure are derived in this article. We also show that the model under conditional independence assumption is nonidentiable subject to right censoring. Our methodology was motivated by and applied to the intensive care unit (ICU) study in Israel.