Statistical models and methods for survival and longitudinal data in clinical and observational studies

点击翻译按钮获取中文摘要

Due to advances in cancer treatment research, more patients are experiencing longer survival and becoming cured. There are increasing demands for appropriate statistical methods to take the cured subjects into account in various complicated sampling designs and models assumptions. Advanced technology also generates big data in various forms, such as tissue microarray data in cancer biology studies. The wide use of observational administrative databases for health service research also creates challenges in causal inferences in the era of big data.******Developing novel statistical models, estimation methods, and accompanying software for analyzing censored survival data and longitudinal data arising in observational and clinical studies has been the primary topic of our research program. We will continue this research program with focus on issues including, but not limited to:******1) Development of model assessment tools and flexible nonparametric models for survival data with a cured fraction******2) Properly modeling correlated survival times and variable selection using marginal survival models and marginal cure models with high-dimensional data******3) Causal inference to determine treatment effects with high-dimensional administrative databases******4) Predictive biomarkers determination for tissue microarray data and joint models for survival and longitudinal data.******5) Development of R packages and SAS macros for proposed models and methods for public use******It is anticipated that this research will generate new statistical methods for statisticians working with high-dimensional clinical and observational data. The new methods will also have impact on other fields of natural science and engineering where event times are collected and analyzed, such as reliability. The research stimulates not only new development in the statistics models and theory, but also knowledge transfer from statistical models to statistical applications due to the development of software for public use to enhance reproducibility and applicability of the proposed research work. The research program will also provide suitable opportunities for trainees at levels from masters students to post-doctoral fellows.