R/node.impurity.R

In gerolt/test2:

node.impurity <- function(			#given
  class.probabilities,			#	a) a table of class probabilities at a node
  impurity.measure = c(	"deviance",	#	b) impurity measure to use when growing tree
                        "gini"))	#the impurity of a node is returned
{
  #evaluate impurity when given class probabilities 
  if (impurity.measure == "deviance") {
    #deviance
    class.probabilities[class.probabilities == 0] <- 1			#change any class.probs from 0 to 1 to make 0*log(0)=0 by default
    impurity <- -2*	sum(	class.probabilities *
                           log(class.probabilities)
    )
  } else if (impurity.measure == "gini") {
    #gini
    impurity <- 1 - sum(class.probabilities^2)
  }
  return(impurity)
}