Heterogeneous ATT: DiD taken locally under forest weights

Usage

did_forest(
  Y,
  D,
  X,
  event_time,
  x_eval = NULL,
  n_trees = 200L,
  min_leaf = 5L,
  alpha = 0.05,
  max_depth = 12L,
  seed = 0L,
  kind = "double-sample",
  clusters = NULL
)

Arguments

Y: Balanced n-by-T panel.
D: 0/1 adoption vector.
X: Covariate matrix.
event_time: H.
x_eval: Optional points at which to evaluate; default training rows.
n_trees: Number of trees.
min_leaf: Minimum leaf size.
alpha: Honest subsample fraction.
max_depth: Maximum tree depth.
seed: Seed for the shared generator.
kind: Forest subsampling scheme passed to hntfst.
clusters: Optional cluster IDs.

Value

A list mirroring the Python RichResult payload.

Examples

# \donttest{
set.seed(4)
n <- 200
X <- matrix(rnorm(n * 2), n, 2)
D <- rbinom(n, 1, 0.5)
Y <- matrix(rnorm(n * 4), n, 4)
Y[, 3:4] <- Y[, 3:4] + D * (1 + X[, 1])
f <- did_forest(Y, D, X, event_time = 3, n_trees = 50L)
names(f)
#>  [1] "estimate"       "tau"            "delta"          "att_uniform"   
#>  [5] "treated_weight" "control_weight" "n"              "n_trees"       
#>  [9] "event_time"     "design"         "method"        
# }