A Short Introduction to Sufficient Dimension Reduction Methods

With technological advances, datasets have grown in both size and complexity. One consequence of increasing amounts of data is that we often need to relate a response variable to a potentially large number of possible covariates. The high dimension of the covariate space makes it difficult to uncover this relationship. Sufficient dimension reduction methods are based on finding a small number of linear combinations of covariates to relate to the response variable. 

In this talk, I am going to give a short introduction to sufficient dimension reduction. I will start by introducing the main ideas of sufficient dimension reduction and the abstract mathematical problem it is trying to solve. After that, I will focus on one or two particular sufficient dimension methods to illustrate how sufficient dimension reduction methods works.