Principal component analysis (PCA) and index construction
Posted: Sun Aug 17, 2014 3:53 pm
Dear guys, I have this problem. I am using PCA for building a sort of leading indicator index using a large dataset. I know that PCA doesn't care about correlation sign. But I think it matters for the correctness of the sign of the component extracted. If my target is to build an economic indicator, I suppose that all the variables have to be positively correlated before extracting the component. Thank you