[R] randomForestSRC VIMP interpretation

Claire's Daddy jerrelharlanmast at gmail.com
Mon Dec 28 19:01:02 CET 2015


For a binary classification model, why variable importance produces
separate importance for each outcome? What does column N and R stand for? I
thought only the first column 'all' is needed. (i.e. randomly permute
variable x, calculated change in out-of-bag prediction error.

*model <- rfsrc(status ~ ., data = breast)*
*x <- predict(model)*
*x$importance *

produces the following

*                              all             N             R
mean_radius          2.880195e-03  2.175676e-03 -2.543478e-03
mean_texture         1.129752e-03  7.837838e-04 -8.043478e-04
mean_perimeter       1.976838e-03  1.466216e-03 -1.652174e-03
mean_area            3.143459e-03  2.054054e-03 -1.782609e-03*


Thanks a lot!


	[[alternative HTML version deleted]]

More information about the R-help mailing list