Hi all,
I am using random forest (regression) and I am having trouble calculating
the variable importance for my object (RF). Here is what I tried:
RF$importance
and
importance(RF)
The thing is that the reported %incMSE is different between those two. I
understand that the function importance() does some re-scaling and averaging
of the results, but does that mean that the order of variables with respect
to importance changes after that?
This is what I got:
> importance(RFcebus)
%IncMSE IncNodePurity
vbio1 28.34672 77950703
vbio4 29.34845 71261614
vbio12 43.62222 126822944
vbio15 34.47683 151702649
vmodis 19.69951 18413449
vfootprint 23.31656 67756166
vgeotopo 30.06409 55710807
vcost_slope 51.85601 210008585
vlight 0.00000 0
vaccess30 182.90648 136540782
> RFcebus$importance
%IncMSE IncNodePurity
vbio1 13171.572 77950703
vbio4 11617.453 71261614
vbio12 18330.536 126822944
vbio15 16786.781 151702649
vmodis 1736.843 18413449
vfootprint 10619.448 67756166
vgeotopo 8619.964 55710807
vcost_slope 25811.859 210008585
vlight 0.000 0
vaccess30 11037.952 136540782
Any suggestions would be appreciatted.
Thanks,
Naiara.
[[alternative HTML version deleted]]
______________________________________________
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
