ABSTRACT
This chapter extends regression analysis beyond linear (continuous) and logistic (dichotomous) outcome to survival analysis with a time-to-event outcome. Survival analysis is used when we want to study a time to an event. In clinical research, analysis of this type of data generally focuses on estimating the probability that an individual will survive for a given length of time. Although measurements of survival times are continuous, their distributions are rarely normal; therefore, the linear regression model assumption is violated. In addition, a common circumstance in working with time-to-event data is that some people in the sample are not observed up to the time of their event (as a result, their data will be censored). Statistical approaches to handling data with such properties are covered in this chapter, including the life table method and product limit method of survival analysis. Kaplan–Meir estimates and curves are discussed along with the log-rank hypothesis test for bivariate analysis.
