Collaborative Research: Conditional Frailty Duration Model for the Study of Repeated Events in the Social Sciences

Repeated events processes, where a subject may experience the same type of event multiple times, are ubiquitous across a great range of important social science applications. The central aim of the research is to develop a robust estimation strategy for assessing covariate effects in repeated events survival models. As the center piece of that strategy, the conditional frailty model combines key features of alternative models designed to account for heterogeneity (a random effect) and for event dependence (stratification), with a conditional definition of the risk set in the same model. The goal of the research is to further test, extend, and apply the conditional frailty model.

The research will advance the application of event history models and enable a deeper understanding of repeated events processes in the social sciences by developing survival estimators that accurately model heterogeneity and event dependence. More specifically, the research looks at the robustness of the conditional frailty model under complications introduced by a) varying degrees of censoring; b) small sample sizes; c) time dependence; and d) alternative forms of event dependence. The conditional frailty model will be extended to deal with a) multiple levels of analysis and b) competing risks. The conditional frailty model will be applied to two central social science questions: What explains the timing and occurrence of a) civil wars and b) the churning of children in foster care? Advances in the development and application of repeated events survival models will be widely applicable in the study of events in the social sciences. Additionally, the methodology will help practitioners in the public policy arena better assess the role of policies and conditions.