Skip to contents

Estimates causal effects under partial/clustered interference, where units within clusters may affect each other's outcomes.

Usage

clustered(
  Y,
  Z,
  X,
  cluster_labels,
  group_labels,
  ingroup_labels,
  cluster_feature = NULL,
  n_moments = 1L,
  prop_idv_model = NULL,
  prop_neigh_model = NULL,
  n_matches = 100L,
  subsampling_match = 2000L,
  categorical_Z = TRUE
)

Arguments

Y

Numeric vector. Outcome variable.

Z

Numeric vector. Treatment indicator (0/1).

X

Numeric matrix. Individual-level covariates.

cluster_labels

Integer vector. Cluster identifier for each unit.

group_labels

Integer vector. Group index within cluster (0-indexed).

ingroup_labels

Integer vector. Position within group (0-indexed).

cluster_feature

Optional numeric matrix. Cluster-level covariates.

n_moments

Integer. Number of moments for covariate aggregation. Default 1.

prop_idv_model

Model wrapper for individual propensity P(Z=1|X).

prop_neigh_model

Model wrapper for neighborhood propensity P(G|X).

n_matches

Integer. Number of matches for variance estimation. Default 100.

subsampling_match

Integer. Maximum subsample size for matching. Default 2000.

categorical_Z

Logical. Treat Z as categorical. Default TRUE.

Value

A clustered object.