This spawned three basic types of response:
1. the decomposition techniques used by SPLUS are "modern" and are less
affected by collinearity than other techniques. Therefore VIF &/or
tolerance measures aren't really necessary.
2. there is a very simple way to get both, expressed elegantly in the
following formula:
VIF<-diag(solve(cor(model.matrix(lm1)[,-1]))), where the matrix term [,-1]
removes the intercept from the correlation matrix.
tolerance is simply the reciprocal of VIF.
Several people sent me variants of this, as well as a function to compute
PRESS statistics.
3. Finally, one writer suggested that this was a design problem, not an
analysis problem involving a
matrices of "near linear dependence" or "near rank loss" best handled by
extracting the "d" matrix from a singular value decomposition of the model
matrix from a lm().
Thanks are due to Bill Venables, Brant Deppa, Frank Harrell, SD Byers,
Stephen Weller, and Michael Conklin for sharing insights with me.
Dr. Marc R. Feldesman
email: feldesmanm@pdx.edu
email: feldesman@ibm.net
pager: 503-870-2515
fax: 503-725-3905
"If ignorance is bliss then why aren't there more happy people?" Lawrence
Peter
Powered by: Monstrochoerus - the 300 MHz Pentium II
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news