8 Principal Component Analysis

Last updated on Sep 27, 2021

Lecture slides: PCA

Today’s exercise: Web Excercise 8

Data sets often include a range of indicators measuring aspects of a latent (i.e., not directly observable) construct, such as intelligence, racism, or homophobia. It has been a big leap forward for the social sciences, when we learned statistical techniques that allow us to identify and measure such latent constructs. In this session I will give a you brief recap of PCA and how to conduct it with R to measure such latent constructs. Below you see a 3D density plot of the two latent constructs we will identify throughout my lecture.

Homework

  1. Read ameoba’s fictional conversation with his grandma about PCA

  2. Finish Exercise 7 and 8.

  3. Think about which of the variables your R script contains are most likely indicators of an underlying latent variable. Go through the ESS codebook and look for additional variables that could further contribute to the latent constructs we have measured (i.e., xenophobia, homophobia, and environmentalism). Update your big PCA.

  4. Once you are done, update your analyses from the past weeks. That is update your analyses (i.e., the figures etc.) so that they now rely on your new scales and indicators.