I'm working with clinical trial data and would like to see whether any cognitive functions (measured by several neuropsychological tasks) changed as a result of the treatment. I'm hoping to derive some composite scores that would represent greater or lesser improvement on any identified principal components.
I have two treatment groups (control and intervention), and each variable is measured twice (once before and once after treatment). Just a regular PCA would involve 4 different covariance matrices (one for each combination of treatment condition and timepoint), but I need a way to pool those covariance matrices. Is this possible/is there freely available R code that would allow me to do this?
Yes, it is possible to conduct a principal component analysis (PCA) on repeated measures data. It would help to combine your pre- and post-treatment scores into a single dataset and then perform PCA on this combined data. Keep in mind PCA doesn't account for the correlation of repeated measures. In your case, repeated measures MANOVA or MANCOVA might be more suitable. These methods already account for the correlation between the repeated measures, eliminating the need to pool covariance matrices.
I add a paper from our research so you can see how we did:
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publication