AHB

A R package for performing matching for observational causal inference on datasets containing discrete and continuous covariates

Documentation here

AHB is a R package for performing matching for observational causal inference on datasets containing discrete and continuous covariates. It implements the mixed integer program algorithm for Adaptive Hyper-Boxes (AHB_MIP_matching and approximate fast algorithm for Adaptive Hyper-Boxes (AHB_fast_matching) which match treatment and control units in unit-specific, hyper-box-shaped regions of the covariate space. The resulting matched groups are interpretable, because the matches are made on covariates, and high-quality, because machine learning is used to determine which covariates are important to match on.

Installation

Dependencies

AHB requires R version (>=4.0.0). Install from here if needed. Required:

• utils (>=4.0.0)
• stats (>=4.0.0)
• Rcpp (>=1.0.4.6)
• bartMachine (>=1.2.5.1) or xgboost (>=1.0.0.2)
• Rcplex (>= 0.3.3) or Rglpk (>=0.6.4)

If your R version does not have these packages, install from here.

User Installation

Download it directly from github