A quick question about the 'Fit Random Forest Model' tool. In the model options, there is an option to use conditional variable importance (party pkg only). The help dialogue states that this option enables conditional interference trees.

As you warn may happen in the help dialogue, ArcMap crashes on me when I select this option (I have 900 records x 9 satellite bands). However, when I run the party pkg without selecting the conditional variable importance box, the summary file still states that the model was fit using conditional interference trees:

So, my question is: does the party package use conditional interference trees after all? I am wondering because my own preliminary research is confirming that the randomForest pkg does indeed do a poor job at estimating variable importance for my data, which is a bit odd as it is all continuous satellite band data.

Thanks again for the help on the several occasions I've reached out!

/Marc