I would like to group principal components based on sample values. That is, for a matrix with columns (PC1, PC2, ... , PCn), and rows with transformed values, I want to group PCs with similar values.

I'm working on a spatial problem, and some PCs produce similar patterns in space. E.g PC3, PC5, and PC18 pick out the same spatial feature, and so these could be grouped.

I have tried hierarchical column-wise clustering to solve this problem. However Euclidean distance doesn't seem appropriate (because PCA produces orthogonal vectors that are ~equidistant?).

I have tried mahalonobis distances with Ward linkage, and produced something that might be appropriate, but need advice from someone with proper mathematical background! How should I group based on PC?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.