Hello, I want to wrote a R function which build an oblique decision tree. The returned tree mix node providing from spliting variables as CART does, with spatial line split. I'm a beginner in R but I have seen that the rpart package provide a "method = usersplits". I think that I can't use it because my spliting criteria need 2 variables (it's for spatial partitionning, so I need space coordinate (x,y). I have already wrote a C function which return the best spliting line. My questions are : - Am I wrong while thinking I can't use rpart with usersplits? - If I'm not, can I use rpart object for drawing my tree (I mean how can I organize my results as an rpart object and use methods for rpart objects)? Guillaume BARBET [[alternative HTML version deleted]]