VCG Sampling using Energy-Based Covariate Balancing

Provides a principled framework for sampling Virtual Control Group (VCG) using energy distance-based covariate balancing. The package offers visualization tools to assess covariate balance and includes a permutation test to evaluate the statistical significance of observed deviations.

eVCGsampler R package

eVCGsampler provides a principled framework for sampling of VCG (Virtual Control Group), using energy distance-based covariate balancing. The package includes visualization tools for assessing covariate balance, as well as a permutation test to evaluate the statistical significance of the deviations.

Example.

Test for 3 covariates before balancing, comparison of the treated groups (TG) with the data pool (POOL), shows high imbalance.

Distance permutation test TG vs POOL:

By running the function: VCG_sampler(treated ~ cov1 + cov2 + cov3, data=dat, n=10)

Distance permutation test TG vs VCG:

Plot specifically for the variable cov3: plot_var(dat_out, what='cov3’)

Best VCG size (exploratory)

With BestVCGsize(treat ~ cov1 + cov2 + cov3, data=dat), you can explore the best size for VCG with the best balance of covariates. It may not necessarily be the best size in terms of power or validity of the study.