Estimating a Continuously Observed Semi-Markov Process

• Activity Number: 236
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Amy Laird*+ and Lurdes Inoue
• Companies: University of Washington and University of Washington
• Keywords: longitudinal study ; chronic disease ; semi-Markov process ; Markov model

Abstract

Longitudinal studies are a powerful method for investigating the natural history of chronic disease. Since many chronic diseases are characterized by a set of health states, a multi-state model is a natural choice. A strong yet commonly used assumption in such models, however, is that the disease process is Markovian. The semi-Markov assumption is much less restrictive, but methods for estimation and inference in semi-Markov models are still being developed. We examine some proposed parametric and nonparametric methods for estimating parameters in a semi-Markov model. We find that while the nonparametric method allows for greater modeling flexibility, it relies on a large number of observed transitions and long observation of the disease process, which may not be a realistic assumption for biomedical studies. In this talk, we will discuss the results from our simulation studies.