ROC score in gbm package

Question

I am having trouble calculating ROC score (AUC) using gbm package. I am working with boosted regression trees. The script I am running is:

testing.tc5.lr005 <- gbm.step(data=ModelData, 
gbm.x = 3:4,
gbm.y = 2,
family = "gaussian",
tree.complexity = 5,
learning.rate = 0.005,
bag.fraction = 0.5)

It should produce "training data ROC score" and "cross validation ROC score" along with other correlation parameters. I do not get ROC scores. I called names(testing.tc5.lr005). It lists cv.roc.matrix but produces the following if I call testing.tc5.lr005$cv.roc.matrix:

[1] 0 0 0 0 0 0 0 0 0 0

I tried calculating ROC AUC by roc(TestData$TN,predTN) (TestData~data used for prediction i.e. observed values; predTN~predicted values). It results in [1]NA. I have no clue what I am doing wrong, although rest of model seems to be doing good with reasonable testing and CV correlation and SE values. Any pointers about where I am wrong or an alternate would be really appreciated! Thanks

Answer 1

First of all, I think you are using the gbm package in conjunction with the dismo package. As Calimo said, gbm.step does not exist outside dismo.

I can't answer your question of HOW to calculate ROC (I'm no expert--I had the same question and that's how I got here), but I can tell you that dismo won't calculate it. I took a look at the code in the dismo package and it appears that they do not allow ROC calculation for the gaussian family.