taddylab / textir Goto Github PK

I have a small matrix with one column containing long strings (1 string per row). If I pass only the one data frame column to tfidf, then I get the error rror in log(nrow(x)) : non-numeric argument to mathematical function. If I pass the whole data frame, I get Error in base::rowSums(x, na.rm = na.rm, dims = dims, ...) : 'x' must be numeric as well as two identical warnings: In Ops.factor(left, right) : ‘>’ not meaningful for factors. I would be very grateful for some insight on what kind of argument this function accepts. thanks!

Hi Matt,

I've been using mnlm for a new text dataset and the predictions don't seem to be working. Here's an example that produces the same error as in my data...

library(textir)
data(we8there)
exes<-we8thereCounts
covars<-we8thereRatings
cl <- makeCluster(detectCores(), type="FORK")
inverse.reg<-mnlm(counts=exes, covars=covars, cl=cl)
PP<-predict(inverse.reg, newdata=exes,type="response")

The error I get is just from this matrix multiplication in predict.dmrcoef:
eta <- t(tcrossprod(t(object[-1,,drop=FALSE]),newdata) + object[1,])
"Cholmod error 'A and B inner dimensions must match' at file ../MatrixOps/cholmod_ssmult.c, line 82"

It works fine with the Cushings and fgl example data, and I've tried monkeying around, but I can't figure out the general solution. Am I just using this wrong or is there a deeper issue?

Any help would be appreciated! I have a dataset that's just begging for a multinomial inverse regression, and the coefficients from the projection functions are already fascinating. But I want to validate the model out-of-sample and this particular error is holding me back.

Thanks!