Print method for wdsmatch objects

Description

Print method for wdsmatch objects

Usage

## S3 method for class 'wdsmatch'
print(x, digits = 4, ...)

Arguments

Value

Invisibly returns the input object x. Called for its side effect of printing a formatted summary to the console, including the point estimate, standard error, confidence interval, number of matches, and sample sizes.

Summary method for wdsmatch objects

Description

Summary method for wdsmatch objects

Usage

## S3 method for class 'wdsmatch'
summary(object, ...)

Arguments

Value

Invisibly returns the input object. Called for its side effect of printing a formatted summary to the console.

Simulated Survey Observational Data

Description

A simulated dataset drawn from a survey-weighted observational study with treatment-dependent (retrospective) sampling. Contains 6 covariates, a binary treatment indicator, observed outcome, and survey design weights. The true PATE is approximately 0.8 and the true PATT approximately 1.0.

Usage

survey_obs

Format

A data frame with approximately 120 rows and 9 variables:

Y

Observed outcome (continuous).

Z

Binary treatment indicator (1 = treated, 0 = control).

X1, X2, X3, X4, X5, X6

Pre-treatment covariates (continuous). The true propensity and prognostic models include an X1:X2 interaction.

survey_weight

Survey design weight (inverse selection probability).

Details

Generated by a simulation where:

• Treatment assignment: P(Z=1|X) = \text{logit}^{-1}(0.3 + 0.6 X_1 + 0.4 X_2 - 0.3 X_3 + 0.2 X_1 X_2).

• Outcome model: Y(0) = 1 + X_1 + 0.5 X_2 - 0.3 X_3 + 0.2 X_4 + 0.3 X_1 X_2 + \varepsilon, with treatment effect \tau(X) = 0.8 + 0.2 X_1.

• Survey selection: treatment-dependent (retrospective) with P(S=1|Z,X) = \text{logit}^{-1}(-2 + 0.3 Z + 0.2 X_1 + 0.15 X_2).

Source

Simulated data; see data-raw/make_survey_data.R.

Examples

data(survey_obs)
head(survey_obs)

# Estimate PATE
fit <- wdsmatchATE(Y = survey_obs$Y, X = survey_obs[, 3:8],
                   Z = survey_obs$Z, weights = survey_obs$survey_weight,
                   M = 3, varest = FALSE)
fit

Weighted Double Score Matching Estimator for Population Average Treatment Effect

Description

Estimates the population average treatment effect (PATE) using weighted double score matching (WDSM) with survey design weights. The method matches treated and control units on arm-specific double scores D_z(X) = (e(X), Psi_z(X)) for z in {0,1}, imputes missing potential outcomes via survey-weighted averaging within match sets, and aggregates using Hajek normalization. Polynomial sieve bias correction removes the finite-sample matching discrepancy. Variance estimation uses a linearization-based multinomial bootstrap that re-estimates score parameters while preserving the original matching structure and survey-weighted reuse frequencies.

Usage

wdsmatchATE(
  Y,
  X,
  Z,
  weights,
  M = 5,
  ps = NULL,
  pg = NULL,
  model.ps = NULL,
  model.pg = NULL,
  sampling = c("retrospective", "prospective"),
  use.bias.correction = TRUE,
  varest = TRUE,
  boots = 200,
  alpha = 0.05
)

Arguments

Details

The estimator achieves double robustness: it is consistent when either the propensity score model or the prognostic score model is correctly specified.

Under retrospective sampling (sampling = "retrospective"), the propensity score is estimated with survey weights to recover the population-level treatment assignment mechanism. Under prospective sampling (sampling = "prospective"), the propensity score is estimated without survey weights. Prognostic scores are always estimated without survey weights, as the conditional outcome mean is invariant to the sampling design.

The sieve basis uses log-odds of the propensity score to match the coordinate system used in matching distance computation.

Value

A list with components:

Examples

data(survey_obs)
fit <- wdsmatchATE(
  Y = survey_obs$Y,
  X = survey_obs[, c("X1","X2","X3","X4","X5","X6")],
  Z = survey_obs$Z,
  weights = survey_obs$survey_weight,
  M = 3,
  model.ps = Z ~ X1 + X2 + X3 + X4 + X5 + X6 + X1:X2,
  model.pg = Y ~ X1 + X2 + X3 + X4 + X5 + X6 + X1:X2,
  sampling = "retrospective",
  varest = FALSE
)
fit

Weighted Double Score Matching Estimator for Population Average Treatment Effect on the Treated

Description

Estimates the population average treatment effect on the treated (PATT) using weighted double score matching (WDSM) with survey design weights. Performs one-sided matching from treated to control units on the control-side double score D_0(X) = (e(X), Psi_0(X)). Only the counterfactual control outcome Y(0) needs imputation; treated outcomes are directly observed. Aggregation uses Hajek normalization over treated-side survey weights. Polynomial sieve bias correction and linearization-based multinomial bootstrap with survey-weighted reuse frequencies are applied. PATT requires only one-sided unconfoundedness: Y(0) independent of Z given X.

Usage

wdsmatchATT(
  Y,
  X,
  Z,
  weights,
  M = 5,
  ps = NULL,
  pg = NULL,
  model.ps = NULL,
  model.pg = NULL,
  sampling = c("retrospective", "prospective"),
  use.bias.correction = TRUE,
  varest = TRUE,
  boots = 200,
  alpha = 0.05
)

Arguments

Value

A list with components:

Examples

data(survey_obs)
fit <- wdsmatchATT(
  Y = survey_obs$Y,
  X = survey_obs[, c("X1","X2","X3","X4","X5","X6")],
  Z = survey_obs$Z,
  weights = survey_obs$survey_weight,
  M = 3,
  model.ps = Z ~ X1 + X2 + X3 + X4 + X5 + X6 + X1:X2,
  model.pg = Y ~ X1 + X2 + X3 + X4 + X5 + X6 + X1:X2,
  sampling = "retrospective",
  varest = FALSE
)
fit