Survival Analysis: From Basic Concepts to Bayesian Models
Details
Speaker: Mick Cooney
Survival analysis is a statistical method for analyzing the expected duration of time until an event occurs. In this one-hour talk, we will walk through the application of survival analysis to life insurance policy lapses, starting from foundational concepts and moving toward Bayesian modeling approaches using Stan.
We will begin by setting the scene with a practical example: the impact of the 2008-2011 credit crisis on Irish life insurance. This provides a backdrop to introduce core survival analysis concepts, including censoring, truncation, survival functions, and hazard functions.
Next, we will review classic approaches to modeling survival data. We will cover the non-parametric Kaplan-Meier estimator, discuss standard parametric models, and examine semi-parametric methods like the Cox Proportional Hazards model.
The second half of the talk will focus on Bayesian survival analysis. Using R and Stan, we will demonstrate how to construct Bayesian models for time-to-event data. We will step through the process of assessing model fit and discuss strategies for improving model performance, focusing on handling right-censored data and quantifying uncertainty.
This session provides an overview for practitioners interested in applying both traditional and Bayesian survival analysis techniques to their own data problems.