Dynamic Modeling of Recurrent Events and Its Applications

University Of North Carolina At Charlotte, Charlotte NC

Investigators

Abstract

This project will develop some novel statistical methods for the recurrent events. The proposed research is motivated by the need to develop new statistical methodology for the malaria vaccine trial and other biometrical researches. The proposed statistical methods can be used to assess the risk factors of event occurrences, how the occurrences of events and event types are affected by concomitant variables, intervention, the past history, as well as the associations among different types of events. The proposed research is challenging in theoretical development and in statistical computation. The proposed statistical methods can facilitate the analysis of the malaria vaccine trial and contribute to develop efficacious malaria vaccines toward control and elimination of malaria. The development of the outlined research would enrich a collection of statistical tools for understanding the event histories and contribute to the efforts to overcome the medical and public health challenges and beyond. The graduate student support will be used for research on the theory of malaria control. The research aims to develop nonparametric and semiparametric dynamic models for the recurrent events that will advance the existing statistical methods for recurrent events. The statistical methods are proposed to model and assess the risk factors of event occurrences, how the occurrences of events and event types are affected by concomitant variables, interventions, and the event history. New statistical methods also proposed to investigate the associations among different types of events. The proposed research is motivated by the challenges arising from analyzing the malaria vaccine efficacy trialand has broach applications in other fields such as biomedical, reliability and economics. The proposed research includes two parts. The research for the dynamic varying-coefficient intensity models of event occurrences is described in Part I which introduces three models: the nonparametric dynamic varying-coefficient intensity model, the semiparametric dynamic varying-coefficient intensity model, and the dynamic models for multiple type of recurrent events. Research for modeling and assessing the associations among different types of recurrent events is proposed in Part II. The nonparametric dynamic varying-coefficient intensity models will be estimated based on the likelihood principle and the local linear smoothing technique. The profile estimation approach will be utilized for the semiparametric dynamic varying-coefficient intensity models. The inference procedures that are of scientific interest will be investigated. Large sample theory will be developed for the proposed estimation and hypothesis testing procedures. A two-stage estimating procedure is proposed to estimate the models for the conditional rate function of the recurrent event processes where the marginal semiparametric models are estimated in the first stage and the models for the rate ratio are estimated in the second stage. Studies are proposed to investigate a variety of models for the conditional rate ratio. New statistical methods, theory and computational algorithms will be developed. The proposed statistical methods can facilitate the analysis of the malaria vaccine trial and contribute to develop efficacious malaria vaccines toward control and elimination of malaria. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.