You've got it right. Just multiply the predicted values by the exposure time or number of cases and that will get you what you want. Note that, depending on the scale of the predictions, you might want to do all the predictions and +log(offset) on the log scale and then exponentiate it. Should be more stable numerically.
Greg ________________________________________ From: Markus Loecher [mailto:markus.loec...@gmail.com] Sent: Tuesday, September 21, 2010 11:15 AM To: r-help@r-project.org Cc: Ridgeway, Greg Subject: package gbm, predict.gbm with offset Dear all, the help file for predict.gbm states that "The predictions from gbm do not include the offset term. The user may add the value of the offset to the predicted value if desired." I am just not sure how exactly, especially for a Poisson model, where I believe the offset is multiplicative ? For example: library(MASS) fit1 <- glm(Claims ~ District + Group + Age + offset(log(Holders)), data = Insurance, family = poisson) head(predict(fit1, data = Insurance, type = "response")) #glm.predict includes the offset: head(predict(fit1, newdata = Insurance, type = "response")) # 1 2 3 4 5 6 # 31.86358 35.27587 28.18080 158.87829 53.97772 84.16012 library(gbm) fit2 <- gbm(Claims ~ District + Group + Age + offset(log(Holders)), data = Insurance, distribution ="poisson", n.trees = 600) head(predict(fit2, newdata = Insurance, type = "response", n.trees=600)) #[1] 0.1378249 0.1378249 0.1314991 0.1284441 0.1389563 0.1389563 #Warning message: #In predict.gbm(fit2, newdata = Insurance, type = "response", n.trees = 600) : # predict.gbm does not add the offset to the predicted values. Would the answer be simple multiplication such as: head(predict(fit2, newdata = Insurance, type = "response", n.trees=600)*Insurance[,"Holders"]) [1] 27.15151 36.38577 32.34878 215.78607 39.46359 74.48058 Any help would be immensely useful. Thx, Markus __________________________________________________________________________ This email message is for the sole use of the intended r...{{dropped:6}} ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
