I have a question about Principal Components Analysis. I've tried to read a lot about it but have a doubt about this application.
I have high dimensional data (known to be similar) which is organized into the columns of matrix Q. I perform a princomp on this data, and project the centered columns of Q onto the calculated eigenvector basis (the first 4).
My question is, can I use the linearly independent vectors of the result A to form a basis for which I can test other high dim data for similarity? e.g. Solve Ax=b for an exact solution? I think the usual way of determining similarity is by Euclidean distance, but I wonder if this is possible too.