Multivariate Analysis 11: tuning loadings, and 3 examples: economics, genetics, and computer vision
After going over strategies for eliminating elements of the loadings matrix in principle component analysis, we try out three different PCA examples. One tries to classify the economies of European nations in the 1970s, another tries to reduce a high-dimensional genetics problem down to just two dimensions, and another is a fun computer vision task that finds the principal components, or "eigenfaces", for a collection of college student ID photos.
This lecture was given in 2021 as part of a Multivariate Analysis module (MT5758) offered to students working toward a variety of one-year master's degree programs related to statistics at the University of St Andrews, in Scotland.