Fit a greedy CI forest from parsnip — ci_forest

A small wrapper used by the ineqTrees parsnip engine for parsnip::rand_forest().

Usage

ci_forest_parsnip(
  formula,
  data,
  weights = NULL,
  rank_name = "wealth",
  outcome_name = "deadu5_num",
  type = "CIg",
  ntree = 500L,
  mtry = NULL,
  minbucket = 100L,
  minsplit = 200L,
  minprob = 0.01,
  maxdepth = 4L,
  min_gain = 0,
  min_relative_gain = 0,
  perturb = list(replace = FALSE, fraction = 0.632),
  na.action = stats::na.omit,
  ...
)

Arguments

formula: A model formula, typically cbind(rank, outcome) ~ predictors.
data: A data frame containing the variables in formula.
weights: Optional case weights supplied by parsnip/hardhat.
rank_name: Name of the socioeconomic rank variable.
outcome_name: Name of the outcome variable.
type: One of "CI", "CIg", "CIc", or "L".
ntree: Number of trees to grow.
mtry: Optional number of variables tried at each split.
minbucket: Minimum weighted child-node size.
minsplit: Minimum weighted parent-node size.
minprob: Minimum child-node weight proportion.
maxdepth: Maximum tree depth. When supplied through set_engine(), this overrides tree_depth.
min_gain: Minimum concentration-index gain required to split.
min_relative_gain: Minimum concentration-index gain as a share of the parent-node impurity required to split.
perturb: Resampling specification passed to ci_forest().
na.action: Function for handling missing values.
...: Additional arguments passed to ci_tree().

Value

A fitted ci_forest object.