AAISM Domain 2 Practice Test

Reducing dimensionality by turning the original variables into a new set of uncorrelated components that preserve as much variability as possible is what Principal Components Analysis does. PCA computes the directions (principal components) that capture the maximum variance in the data and makes them orthogonal, so each component is uncorrelated with the others. The components are ordered by the amount of variance they explain, with the first component describing the most variance, the second describing the next most, and so on. Because of this setup, you can often keep only the first few components to reduce dimensionality while retaining most of the information in the data.

This differs from factor analysis, which seeks latent factors to explain correlations among variables and does not necessarily maximize explained variance with orthogonal components. K-means clustering groups data into clusters rather than creating new descriptive components. Linear regression focuses on predicting a response variable and does not inherently transform predictors into a parsimonious, uncorrelated set of components.