Author: Tim Foden
Date: 23:29:33 04/05/04
Go up one level in this thread
On April 05, 2004 at 11:51:57, Roger D Davis wrote: >Tim, > >If you're using multivariate regression, doesn't that assume that the regression >variables are normally distributed? > >Roger I'm not really sure. My maths is OK, but not great. As far as I know it's just doing a 'least squares' best fit. AFAIK, this should work whatever the distribution, but it can intruduce large errors at the 'edges' of the hyper-plane. This whole thing was just a test. I don't guarantee the accuracy of any of the values... as I said I know I still have potential problems with the method. I felt that this was an interesting first test to do though. Cheers, Tim.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.