447-6236-00L  Statistics for Survival Data

SemesterSpring Semester 2020
LecturersA. Hauser
Periodicitytwo-yearly recurring course
Language of instructionEnglish


AbstractThe primary purpose of a survival analysis is to model and analyze time-to-event data; that is, data that have as a principal endpoint the length of time for an event to occur. This block course introduces the field of survival analysis without getting too embroiled in the theoretical technicalities.
ObjectivePresented here are some frequently used parametric models and methods, including accelerated failure time models; and the newer nonparametric procedures which include the Kaplan-Meier estimate of survival and the Cox proportional hazards regression model. The statistical tools treated are applicable to data from medical clinical trials, public health, epidemiology, engineering, economics, psychology, and demography as well.
ContentThe primary purpose of a survival analysis is to model and analyze time-to-event data; that is, data that have as a principal endpoint the length of time for an event to occur. Such events are generally referred to as "failures." Some examples are time until an electrical component fails, time to first recurrence of a tumor (i.e., length of remission) after initial treatment, time to death, time to the learning of a skill, and promotion times for employees.

In these examples we can see that it is possible that a "failure" time will not be observed either by deliberate design or due to random censoring. This occurs, for example, if a patient is still alive at the end of a clinical trial period or has moved away. The necessity of obtaining methods of analysis that accommodate censoring is the primary reason for developing specialized models and procedures for failure time data. Survival analysis is the modern name given to the collection of statistical procedures which accommodate time-to-event censored data. Prior to these new procedures, incomplete data were treated as missing data and omitted from the analysis. This resulted in the loss of the partial information obtained and in introducing serious systematic error (bias) in estimated quantities. This, of course, lowers the efficacy of the study. The procedures discussed here avoid bias and are more powerful as they utilize the partial information available on a subject or item.

This block course introduces the field of survival analysis without getting too embroiled in the theoretical technicalities. Models for failure times describe either the survivor function or hazard rate and their dependence on explanatory variables. Presented here are some frequently used parametric models and methods, including accelerated failure time models; and the newer nonparametric procedures which include the Kaplan-Meier estimate of survival and the Cox proportional hazards regression model. The statistical tools treated are applicable to data from medical clinical trials, public health, epidemiology, engineering, economics, psychology, and demography as well.