Skip to contents

Run the full doubly-robust estimator panel

Source: R/panel.R
dr_panel.Rd

Fits the shared Super Learner nuisances under all three schemes (full-sample, cross-fit, double cross-fit) and runs every estimator on them, returning the eight-row panel that is the organizing idea of Three Routes to One Answer: the singly-robust baselines (g-computation, IPTW) plus the six doubly-robust estimators (AIPW, TMLE, DML, CV-TMLE, DC-AIPW, DC-TMLE), all on identical nuisances so any difference is estimator-driven.

Usage

dr_panel(
  Y,
  A,
  X,
  family_y = stats::gaussian(),
  y_range = NULL,
  V = 5L,
  Vd = 3L,
  SL.library = default_dr_library,
  seed = 1L,
  truth = NA_real_
)

Arguments

Y

Numeric outcome vector.

A

Binary treatment vector (0/1).

X

Covariates (data frame or matrix).

family_y

Outcome family: stats::gaussian() (continuous) or stats::binomial() (binary).

y_range

Optional outcome range for the TMLE scaling; inferred as c(0, 1) automatically when Y is binary.

V

Number of folds for "crossfit".

Vd

Number of folds for "double".

SL.library

Super Learner library shared by g and Q (default default_dr_library).

seed

RNG seed for fold construction.

truth

Optional known ATE. When supplied, a bias column is filled in (useful on semi-synthetic benchmarks such as IHDP or Twins).

Value

A data.frame of class dr_panel with one row per estimator x nuisance strategy (estimator, nuisance, estimate, se, ci_lo, ci_hi, bias), carrying truth, n, and call attributes.

See also

dr_nuisance(), dr_aipw(), dr_tmle(), plot.dr_panel()

Examples

if (FALSE) { # \dontrun{
d <- dr_nhefs()
fit <- dr_panel(d$Y, d$A, d$X)
fit
plot(fit)
} # }