Estimate a causal effect through the full MRM pipeline — morie_mrm_estimate_causal

The estimation step of the MRM: runs the requested native estimators (nearest-neighbour matching, IPW/design-based ATE, AIPW, DML) on the same specification, applies a multiple-testing correction across them, and bundles estimates, corrected p-values, confidence intervals, per-method diagnostics, and a citation block into one results object.

Usage

morie_mrm_estimate_causal_effect(
  data,
  treatment,
  outcome,
  covariates,
  methods = c("matching", "ate", "aipw", "dml"),
  correction = "holm",
  seed = 42L
)

Arguments

data: Data frame (or morie_mrm_dataset / morie_mrm_reconciliation, whose matched rows are used).
treatment, outcome: Column names (binary 0/1 treatment).
covariates: Character vector of adjustment covariates.
methods: Subset of \(c("matching", "ate", "aipw", "dml")\) (default all four).
correction: Multiple-testing correction passed to stats::p.adjust (default "holm").
seed: RNG seed forwarded to the stochastic estimators.

Value

An object of class morie_mrm_effect: list with results (data frame: method, estimate, std_error, ci_lower, ci_upper, p_value, p_adjusted), consensus (inverse-variance pooled estimate; its standard error is the weighted mean of the standard errors, because the estimators share one data set and are not independent), correction, spec, citation.

Examples

set.seed(1)
n <- 400
x <- rnorm(n)
t <- rbinom(n, 1, plogis(0.5 * x))
y <- 1 + 0.8 * t + 0.5 * x + rnorm(n)
df <- data.frame(y = y, t = t, x = x)
eff <- morie_mrm_estimate_causal_effect(df, "t", "y", "x",
                                        methods = c("ate", "aipw"))
eff$results$estimate
#> [1] 0.8523738 0.8585237