Trains an instrumental forest that can be used to estimate conditional local average treatment effects tau(X) identified using instruments. Formally, the forest estimates tau(X) = Cov[Y, Z | X = x] / Cov[W, Z | X = x]. Note that when the instrument Z and treatment assignment W coincide, an instrumental forest is equivalent to a causal forest.
instrumental_forest(X, Y, W, Z, Y.hat = NULL, W.hat = NULL,
Z.hat = NULL, sample.weights = NULL, sample.fraction = 0.5,
mtry = NULL, num.trees = 2000, min.node.size = NULL,
honesty = TRUE, honesty.fraction = NULL, ci.group.size = 2,
reduced.form.weight = 0, alpha = 0.05, imbalance.penalty = 0,
stabilize.splits = TRUE, clusters = NULL,
samples.per.cluster = NULL, compute.oob.predictions = TRUE,
num.threads = NULL, seed = NULL)The covariates used in the instrumental regression.
The outcome.
The treatment assignment (may be binary or real).
The instrument (may be binary or real).
Estimates of the expected responses E[Y | Xi], marginalizing over treatment. If Y.hat = NULL, these are estimated using a separate regression forest.
Estimates of the treatment propensities E[W | Xi]. If W.hat = NULL, these are estimated using a separate regression forest.
Estimates of the instrument propensities E[Z | Xi]. If Z.hat = NULL, these are estimated using a separate regression forest.
(experimental) Weights given to each observation in estimation. If NULL, each observation receives equal weight.
Fraction of the data used to build each tree. Note: If honesty = TRUE, these subsamples will further be cut by a factor of honesty.fraction.
Number of variables tried for each split.
Number of trees grown in the forest. Note: Getting accurate confidence intervals generally requires more trees than getting accurate predictions.
A target for the minimum number of observations in each tree leaf. Note that nodes with size smaller than min.node.size can occur, as in the original randomForest package.
Whether to use honest splitting (i.e., sub-sample splitting).
The fraction of data that will be used for determining splits if honesty = TRUE. Corresponds to set J1 in the notation of the paper. When using the defaults (honesty = TRUE and honesty.fraction = NULL), half of the data will be used for determining splits
The forst will grow ci.group.size trees on each subsample. In order to provide confidence intervals, ci.group.size must be at least 2.
Whether splits should be regularized towards a naive splitting criterion that ignores the instrument (and instead emulates a causal forest).
A tuning parameter that controls the maximum imbalance of a split.
A tuning parameter that controls how harshly imbalanced splits are penalized.
Whether or not the instrument should be taken into account when determining the imbalance of a split.
Vector of integers or factors specifying which cluster each observation corresponds to.
If sampling by cluster, the number of observations to be sampled from each cluster when training a tree. If NULL, we set samples.per.cluster to the size of the smallest cluster. If some clusters are smaller than samples.per.cluster, the whole cluster is used every time the cluster is drawn. Note that clusters with less than samples.per.cluster observations get relatively smaller weight than others in training the forest, i.e., the contribution of a given cluster to the final forest scales with the minimum of the number of observations in the cluster and samples.per.cluster.
Whether OOB predictions on training set should be precomputed.
Number of threads used in training. By default, the number of threads is set to the maximum hardware concurrency.
The seed of the C++ random number generator.
A trained instrumental forest object.