Chee Chen
2011-Apr-19 01:34 UTC
[R] How to generate orthogonal matrix with less computational cost
Dear All, I would like to ask a question on how to generate an orthogonal matrix with less computational complexity. Basically I know two ways of getting an orthogonal matrix: 1. via the QR decomposition 2. via the SVD Both are essentially the same. A friend of mine told me to use package Matrix, but I did not know where to start. Question: However, is there a way to generate such a matrix without resorting to these methods of great computational costs, since I will have to generate very large dimensional orthogonal matrix and correlation matrix (say, dimension = 10^6)? Please forgive me if 10^6 is not big enough. Your help is greatly appreciated! Regards, -Chee P.S. Thanks to Phil Spector and I got it! [[alternative HTML version deleted]]