Julie Lee-Yaw
2015-Jul-13 11:34 UTC
[R] Party package: varimp(..., conditional=TRUE) error: term 1 would require 9e+12 columns (fwd)
Hello, I'm following previous threads on the use of varimp in the R "party" package. I'm running up against similar problems to that described here (and the thread referenced therein): https://stat.ethz.ch/pipermail/r-help/2011-October/292897.html Specifically, I'm running cforest and then varimp on a dataset with one binary (yes/no) variable to classified and 32 predictors (This is in the context of using environmental data to predict site occupancy and all my predictors are numeric or integer). As described in previous posts, "varimp" command is giving me the following error: "Error in model.matrix.default(as.formula(f), data = blocks) : term 1 would require 1e+25 columns" (I've tried changing the threshold parameter with no luck).? It seems like one of the package author's responses (and possible solutions) to a previous inquiry may have been made outside of the listserv and I'm wondering if it could be posted for reference? Is there something wrong with applying this method given the nature of my data (e.g. many, strongly correlated predictors)? Thanks in advance! [[alternative HTML version deleted]]