BAYESIAN ANALYSIS OF DURATION DATA USING MULTISTATE COMPETING RISKS MODELS

Getachew A. Dagne (U. S. A.)

Received July 19, 2007

Abstract

This paper presents the results of multistate competing risks modeling of social interaction between parents and children. A multilevel model, based on continuous-time semi-Markov chains, is proposed to adequately characterize the behavioral processes of these dyads. The model deals with repeated events within dyads in which there are multiple origin states and multiple transition states. The transitions are jointly modeled to allow interdependence among competing risks. The proposed method is applied to real observational data to assess the relation of emotional expression in parent-child interaction to risk for early and persisting child conduct problems.

Keywords and phrases: competing risks, event history, survival, multilevel models, multistates, Bayesian inference, semi-Markov models.