I have built several spatial models for a binary outcome,
I got a very strange prediction for the GBM (package gbm) as it seems that there are predictions outside the study area. I initially thought that these were NAs but I checked it and there are no NAs.
This is my code
formula_GBM <- as.formula(paste("presence ~", paste(preds_selected, collapse = "+"))) mod_GBM <- gbm(formula_GBM, data = dataset, distribution="bernoulli") GBM_P <- predict(preds_cut, mod_GBM, type = "response")
when I plot the predictions along with the other models I can see that something is wrong
Does anyone have any hints for this behaviour?