Estimate per-unit Conditional Average Treatment Effects (CATE)

The T-learner fits separate outcome models on treated and control units, then predicts the counterfactual for each unit: \(\widehat{CATE}_i = \hat{\mu}_1(X_i) - \hat{\mu}_0(X_i)\).

Usage

morie_estimate_cate(
  data,
  treatment,
  outcome,
  covariates,
  propensity_col = NULL,
  outcome_model = c("linear", "logistic"),
  meta_learner = c("t_learner", "s_learner", "x_learner", "dr_learner")
)

Arguments

data: A vector; indexed elementwise.
treatment: Passed to morie_estimate_propensity_scores.
outcome: See Usage.
covariates: Passed to morie_estimate_propensity_scores.
propensity_col: Optional; may be NULL. Passed to is.null.
outcome_model: Family for the outcome model: "auto" (logistic for a 0/1 outcome, linear otherwise), "linear" or "logistic".
meta_learner: One of "t_learner" (default), "s_learner", "x_learner" (Kuenzel et al. 2019 – two-stage with propensity-weighted combination; strong under arm imbalance), or "dr_learner" (Kennedy 2023 – cross-fit AIPW pseudo-outcome regressed with the native forest; doubly robust). All four are native; outcome_model applies to the T/S-learners only.

Value

Numeric vector of per-unit CATE estimates.

Details

The S-learner fits one model with treatment as a feature.

For random-forest CATE estimation prefer grf::causal_forest (richer heterogeneity, honest sample splitting).

Examples

set.seed(1)
morie_estimate_cate(
  data = data.frame(
    t = stats::rbinom(100, 1, 0.4),
    y = stats::rbinom(100, 1, 0.3), x1 = stats::rnorm(100),
    x2 = stats::rnorm(100)
  ), treatment = "t", outcome = "y",
  covariates = c("x1", "x2")
)
#>   [1]  0.024465332  0.039356269 -0.040246877  0.117055146  0.111672149
#>   [6]  0.162579814  0.132324461  0.153387263  0.126130614  0.221687262
#>  [11]  0.042330327  0.037914880  0.230386067  0.081385490  0.041083866
#>  [16]  0.007127270  0.074669018  0.109843821  0.113769950  0.078260489
#>  [21]  0.106144561  0.209144017  0.094614811  0.087982177  0.122194096
#>  [26]  0.093948665  0.096886758  0.142669661  0.026228031  0.055154705
#>  [31]  0.130448792  0.137599893  0.159952058 -0.059160971  0.120918303
#>  [36] -0.038353437  0.047222047  0.085900174 -0.001304765  0.087434886
#>  [41] -0.098235321  0.160103194 -0.060315776  0.081240777 -0.043927473
#>  [46]  0.093747288  0.292275959  0.102285354 -0.014286989 -0.084867519
#>  [51]  0.126872808  0.076968214  0.066834626  0.019537909 -0.060651204
#>  [56] -0.039819205  0.259763195  0.018920114 -0.040937296  0.269294722
#>  [61]  0.097416770  0.084449855  0.193941856  0.140738997 -0.019289199
#>  [66]  0.275365028  0.084125581  0.007912745  0.090712434  0.141748531
#>  [71]  0.301945115  0.088230399  0.160536609 -0.003174590  0.057893709
#>  [76]  0.051360169  0.237445731  0.248202738  0.225770837  0.166657852
#>  [81] -0.028382634  0.191567314  0.067050019 -0.012141101  0.157035910
#>  [86]  0.111887646  0.242167925 -0.006161392  0.040217899 -0.022109940
#>  [91]  0.089840862  0.114638264  0.020292394  0.212288985 -0.072555701
#>  [96] -0.003585252  0.192139157 -0.028253494  0.129795491  0.069249681