2

Hack-R's answer (the first one) to this question:

How to actually plot a sample tree from randomForest::getTree()?

shows a plot for a single binary decision tree in a random forest. I was wondering what the "p<0.001" represents?

Community
  • 1
Learning_R
  • 43
  • 4

1 Answers1

3

It's because the random forest algorithm being used is based on a conditional inference framework (rather than Breiman's original algorithm). It essentially determines whether a split should be carried out by using a p-value calculated from a permutation test, corrected for multiple comparisons. The main benefit of this approach over Breiman's is that conditional inference forests yield unbiased variable importance metrics. It's available in R using the party package, and you can read more about it here.

dmartin
  • 3,010
  • 3
  • 22
  • 27