utree_control: Control for uplift trees.

In leoguelman/uplift2: Uplift Modeling

Control for uplift trees.

Description

Various parameters that control aspects of the utree fit.

Usage

utree_control(minsplit = 40L, minbucket.t = 20L, minbucket.c = 20L,
  var.select.criterion = "pvalue", var.select.test = "asymptotic()",
  alpha = 0.05, bonferroni = FALSE, balance.sample = "undersample",
  split.criterion = "uplift", maxdepth = Inf, mtry = Inf)

Arguments

minsplit	The minimum number of observations in a node in order to be considered for splitting.
minbucket.t	The minimum number of treatment observations in any terminal <leaf> node. The treatLevel can be used to determine the treatment level of interest.
minbucket.c	The minimum number of control observations in any terminal <leaf> node.
var.select.criterion	The criterion used to select the variable for splitting. At the moment, only "pvalue" is accepted. The variable with minimum pvalue is selected for splitting.
var.select.test	The conditional null distribution of the test statistic. This is passed to the distribution argument in coin::independence_test. For example, for an approximative (Monte Carlo) reference distribution with B Monte Carlo replicates, use approximate(B=999).
alpha	The maximum acceptable pvalue required in order to make a split.
bonferroni	Apply bonferroni adjustment to pvalue?
balance.sample	The sampling method used to balance the treatment variable. This attempts to have an equal representation of each treatment before implementing the independence test described in var.select.test. The options are "undersample" (default), "oversample", "none". See the argument sampling in mom for details.
split.criterion	The split criteria used at each node of each tree; possible values are: "uplift" (default), "kld" (Kullback-Leibler divergence), "ed" (Euclidean divergence), or "l1d" (L1-norm divergence). See details in Guelman et al. (2015).
maxdepth	Maximum depth of the tree. The default maxdepth = Inf means that no restrictions are applied to tree sizes.
mtry	Number of input variables randomly sampled as candidates at each node. The default mtry = Inf means that no random selection takes place.

Value

A list.

Author(s)

Leo Guelman [e-mail]

leoguelman/uplift2 documentation built on April 15, 2022, 4:34 a.m.