Dimension Reduction and Classification Using PCA, Factor Analysis and Discriminant Functions - A Short Overview
Course Topics
Tuesday, October 28: Often researchers are faced with data in very high dimensions (e.g. too many predictors for a regression model), or must come up with a rule to classify data in pre-determined groups.
This course will cover statistical approaches to dimension reduction using principal components and factor analysis, and classification using discriminant functions. The first part will outline the geometric concept behind principal components, its application in a typical problem followed by methods on how to decide how many components to choose. The second part will cover factor analysis as a concept and a typical application with interpretation of factors. The last part will focus on the concept of classification and use of discriminant functions as a specific tool to classify ages.
Statistical collaborators from the LISA will be present at the end of the lecture to answer questions on how a specific researcher's data can be thought of as one of these problems. The examples will mostly be done using JMP or R.