Calculates the average minimal depth of each predictor used to fit
a CERFIT object. It calculates Variables importance by using a Variables average minimal depth.
variable's with a lower average minimal depth are more important.
Usage
MinDepth(cerfit)
Value
Returns a named vector with the name of each predictor used to fit the CERFIT
object and its corresponding average minimal depth across all trees
Arguments
cerfit
A fitted CERFIT object
Details
The depth of the root node is zero and if a variable does not appear
at any split in a tree it is assigned maxdepth + 1 for that tree.
fit <- CERFIT(Result_of_Treatment ~ sex + age + Number_of_Warts + Area + Time + Type | treatment,
data = warts,
ntrees = 30,
method = "RCT",
mtry = 2)
importance <- MinDepth(fit)