Approaches to Survival Analysis with Machine Learning

This talk provides a concise introduction to survival analysis using time-to-event data. A continuous analogue of the traditional Kaplan-Meier estimator is derived, which can be used to perform survival analysis with modern machine learning algorithms. Survival analysis is a leading example of a practically important statistical method that is not simply a prediction/forecasting problem on account of the training data being unavoidably censored. Specifically, in this talk I will emphasize the distinction between an estimator and an estimand as it relates to reduced form versus structural (latent variable) models.

Bio
https://math.la.asu.edu/~prhahn/