EViews.com

Hi all,

I am working on a principal component analysis (PCA) on a group of variables, and I find that the component scores are not a linear combination of my original data variables (with the eigenvector as the loadings). Is this because Eviews will transform the original data when performing PCA? If so, how can I express the component scores in terms of the original data and variables?

Thanks!

MUS

By default the scores are computed for correlation matrix of the series in the group. If you want the scores in the original scaling, you'll have to compute them using the covariance matrix (you may even have to center the variables first for them to be linear combinations of the original data, I can't remember off the top of my head).