Singular Value Decomposition

[blonggovnl]

Hi;

I'm trying to implement a Lee-Carter Model in EViews, but am having an issue with singular value decomposition. The issue is related to the dimensions of the matrices involved.

Code: Select all

vector s matrix v matrix u=@svd(y_m,s,v)

I know EViews automatically resizes s and v to the appropriate dimensions, and they look fine. My issue is with u. The matrix y_m is 91 x 42 and as such, u should be 91 x 91, but I am getting a u that has the same dimensions as y_m.

Thoughts?

[EViews Gareth]

Yes, EViews only computes the first n columns of u (called the "economy SVD" in packages such as Matlab).

If you want the full matrix, you can always compute the eigenvectors directly.