2

I'm trying the gbm.step function from dismo package to fit a gbm model using cross-validation.

According to the dismo document(https://cran.r-project.org/web/packages/dismo/dismo.pdf, page 32), the return value of gbm.step is a gbm object. The gbm object has a "cv.fitted" value which refers to the cross-validation predicted values (https://cran.r-project.org/web/packages/gbm/gbm.pdf, page 15).

I want to obtain the "cross-validation predicted values", but I found the return value from gbm.step function only contains "fitted" not "cv.fitted".

This document (https://cran.r-project.org/web/packages/dismo/vignettes/brt.pdf, page 5) says "fitted" is the fitted values from the final tree, on the response scale.

I'm not sure can I use this "fitted" to say it is the "cross-validation predicted values"? Otherwise, how can I get the cv predicted value from gbm.step?

bouncyball
  • 10,631
  • 19
  • 31
hiyu
  • 21
  • 2
  • Although I found the source codes (https://brage.bibsys.no/xmlui/bitstream/handle/11250/196837/SD_analysis.txt?sequence=14) of a paper which used the same way "fitted" from the gbm model to obtain the predicted value, I'm still not sure enough on this... – hiyu Jan 14 '18 at 23:41

0 Answers0