Survival Analysis and Related Topics

点击翻译按钮获取中文摘要

Ying The research develops statistical models, methodologies and related theory to address issues arising from design, modeling and analysis of survival data, which are often subject to censoring, truncation and/or other types of missingness. It introduces new hazard-based regression models for survival data that extend and complement the standard proportional hazards regression, and yet still possess essential good features that the latter has. Besides the usual right-censored data, it also provides solutions to such important and difficult problems as interval censoring, random effects and multivariate survival times by exploiting model variation and developing suitable analytic tools. It involves finding computationally convenient representations and establishing sharp theoretical properties for the analysis of discrete (survival) data, particularly of those that can be summarized by two-by-two or more general contingency tables. New methods and theory tailored to retrospective studies, which are often summarized as the case-control and case-cohort regression problems, are also developed. Survival data are collected in many scientific investigations, including biomedical follow-up studies, industrial life tests, study of economic durations, astronomy, demography among others. Development of suitable statistical methods is crucial to handling of such data so that valid scientific conclusions can be derived. This project develops several novel approaches, some of which are suitable for identification of health related risk factors, such as exposure to hazardous environment, consumption of certain foods etc., and others are applicable to evaluation of therapeutic effects as well cost effectiveness of medical treatments. These new methods may impact other areas of public and scientific interests, an example of which is the study of factors influencing unemployment duration.