Skip to contents

Internal wrapper for policy tree computation

Source: R/margot_policy_tree_utils.R
dot-compute_policy_tree.Rd

Internal function that wraps either policytree::policy_tree() or fastpolicytree::fastpolicytree() depending on user preference and package availability.

Usage

.compute_policy_tree(
  X,
  Gamma,
  depth,
  tree_method = "policytree",
  min_node_size = NULL
)

Arguments

X

Covariate matrix

Gamma

Matrix of doubly robust scores

depth

Integer depth of tree (1 or 2)

tree_method

Character string: "policytree" or "fastpolicytree"

min_node_size

Integer or NULL. Smallest permitted terminal node size.

Value

Policy tree object

Details

Margot pins fastpolicytree's strategy.datatype to 1, which uses unsorted data sets that are sorted on demand. The upstream automatic strategy (2) can select a different, lower-value rule than exact policytree for wide covariate matrices. The pinned strategy is therefore part of Margot's fast-engine contract rather than an optimisation option selected from the input data.