Professor: Prof. Grace Yang

Prerequisites: Minimum background in Stat 410 and Stat 700 or equivalent (with consent of the instructor). 

Survival analysis concerns the statistical theory and methods for the analysis of time-to-event data or lifetime data. Lifetime data are commonly studied by researchers from diverse scienti c elds including physics, biology, medicine, public health, actuaries, epidemiology, economics and engineering reliability. Measure of lifetime requires the knowledge of both its beginning and its end points. These end points are not always possible to measure. For example, typically a detector can record the time at which a neutron disintegrates, but not the time it is generated, and this results in a left-censored neutron lifetime. Just the opposite in clinical trials, the survival time of a patient after a treatment might be right-censored if the patient withdraws from the trial before his/her death or if the trial terminates before the patient dies.  There are numerous other forms of incomplete lifetimes such as doubly censored, interval censored, randomly truncated and current status data. Sampling methods, sampling subjects, experimental designs and limitations in recording instruments are among the contributing factors to the incomplete data. Proper treatment of incomplete data is necessary for eliminating bias in data analysis. The course is mainly about the statistical analysis of incomplete lifetime data. The topics to be presented include stochastic modeling of censored and random-truncation data, parametric and nonparametric methods, the Kaplan-Meier estimator of a survival function, the Lynden-Bell estimator, construction of con dence bands, the Cox regression model, logistic models, the Fix-Neyman competing risks model, asymptotic statistical inference. Emphasis will be on statisical methods with examples drawn from applications.