STAT 425 Week 10 - Shrinkage Methods for Regularized Regression - Part 1

Discusses principal component regression as a way to both understand the variation in the regression predictor variables, and potentially reduce the dimension of the regression model. The idea is to use the first few orthogonal projections of the columns of the design matrix, projections accounting for the the majority of variation, as new predictors for the response variable. We can optimize this procedure using selection criteria such as cross-validation prediction error.