Skip to content
Prev 2983 / 7420 Next

how-to identify redundant predictors

On Sat, Apr 21, 2012 at 12:02 PM, C Hess <13184 at stud.leuphana.de> wrote:
You have not mentioned how many samples and predictor variables you
have, but I think that you can also use PLS (partial least square
regression, package "pls"), especially if you have many, possibly
correlated, predictor variables, and relatively few samples. The
accompanying web-site (http://mevik.net/work/software/pls.html)
provides R-code that implements the VIP (variance importance in
projection) algorithm that might be useful during variable selection.

Cheers,
Ivailo