wdsmatch - Weighted Double Score Matching for Survey-Weighted Causal Inference

Implements weighted double score matching (WDSM) for estimating population-level causal effects from complex survey data. Combines propensity scores and prognostic scores with survey design weights for matching, survey-weighted imputation within match sets, and Hajek normalization to target the population average treatment effect (PATE) and the population average treatment effect on the treated (PATT). Supports both retrospective (treatment-dependent) and prospective (treatment-independent) sampling designs. Uses propensity probabilities and arm-specific prognostic scores for matching, with a complete quadratic bias correction in each arm's double score. Provides linearization-based multinomial replication variance estimates and centered normal Wald confidence intervals, retaining the original matching reuse coefficients without re-matching. Supplied scores can be held fixed for inference conditional on those scores. This weight-only interface does not encode survey strata, clusters, or design-specific replicate weights. The method is described by Zeng et al. (2026) <doi:10.48550/arXiv.2610.25870>.