I am trying to use H2O in R to run a random forest. http://docs.h2o.ai/h2oclassic/Ruser/rtutorial.html

In the documentation, I saw that there is an option for an offset parameter but I cannot find much information about how it is leveraged.

In logistic regression, I have used an offset in two ways: 1. To adjust for oversampling a binary event (http://support.sas.com/kb/22/601.html) 2. To do a two stage model where the first stage logit is calculated and then I used the logit score as an offset in the second stage so in effect the residual is modeled in stage two.

I would like to replicate both of these through random forest, if possible, but did not think it was possible until I saw the offset parameter in the H2O implementation of Random Forest. Does anyone know if the H2O offset parameter functions the same as the offset option in SAS proc logistic?


thanks for pointing this out! The offset parameter is actually not supported for H2O's distributed random forest. The parameters will be remove in a future release. A jira ticket for the issue can be found here:https://0xdata.atlassian.net/browse/PUBDEV-5191


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.