I've implemented Random Forest regression algorithm in R (randomForest package) and GEE, but they are giving me very different results (average difference is 4% o and going up to 19% in some cases), and mainly the extreme values (low and high) differ very much.
I'm controlling for: number of trees variables per split seed node size. And this is all the common parameters between the two implementations, as far as I understand.
What else can I do to improve the agreement between the two? I did not found any useful documentation of the GEE implementation.