The class on Survival Analysis used the book Survival Analysis by John P Klein and Melvin L Moeschlberger. The class covered ideas from survival analysis starting with the survival function, hazard function, mean residual life function, and common parametric models for survival data. Regression models for survival data modeled the survival time for a variable X off of a group of covariates Z where Z may include treatment, risk features, ad confounders. One model was the accelerated failure time model. Another model discussed more in depth was the proportional hazards regression model. Finally we briefly touched in 255 upon the stratified proportional hazard model and additive hazards models.
The class also involved dealing with the problem of censoring and truncation and what the difference between the two. In particular, for truncation we selectively sample within a given period and therefore do not have any data outside of a time horizon while in censoring we simply have incomplete observation of the data. Censoring could be further divided into types of censoring. For instance, there is right censored data and left censored data (Left censored data can often happen when we are looking for “first use” type of data and we don’t know when the first use happens). Right censored data could be censored after a given period of time a patient is going to be followed or a study could have a defined end date based on the calendar end time of a study.
Truncation and Censoring have to be dealt with to accurately model the distribution of survival of events for subjects. The class then involved different methods for confidence intervals for cumulative hazard functions given by pointwise confidence intervals and confidence bands. The class finally discussed model selection using diagnostic criterion (chi-squared tests), and culminated in a final project using Mayo Clinic data. Statistics 293 further pursued the topic of Survival Analysis and explored the idea of time-varying coefficient models and joint modeling in survival analysis models.