Composite Rubin-style identifiability assumption check

Returns each assumption with diagnostic evidence + a flag.

Usage

mrm_assumptions_check(data, treatment_col, outcome_col, covariates)

Arguments

data

data.frame.

treatment_col

Binary 0/1 column.

outcome_col

Outcome column (presently unused; reserved for future E-value evidence).

covariates

Character vector of covariate columns.

Value

Named list with sutva, unconfoundedness, probabilistic_assignment, overall_verdict sub-lists.

Examples

set.seed(2026)
n <- 300L
x <- rnorm(n)
D <- rbinom(n, 1, plogis(0.5 * x))
y <- 0.7 * D + 0.3 * x + rnorm(n)
df <- data.frame(D = D, y = y, age = x)
chk <- mrm_assumptions_check(df,
  treatment_col = "D",
  outcome_col = "y",
  covariates = "age"
)
chk$overall_verdict
#> [1] "one or more diagnostic flags; see fields"