Discussion questions#
Principal component regression (PCR) and partial least squares (PLS) both work by effectively reducing the complexity (i.e., variance) of a model by accounting for correlations across variables. How is this different from ridge regression, which also accounts for correlational structure in the data set to manage model complexity?
When would methods like PCR or PLS be preferred over subset or stepwise selection methods? When would subset or stepwise selection methods be preferred over PCR or PLS?