VsusP: Variable selection from Bayesian shrinkage priors in R and Python

Bayesian shrinkage regression estimates many coefficients jointly, but continuous priors do not usually produce exact zeros. Researchers therefore need a separate rule to select variables from fitted coefficient draws. VsusP provides R and Python interfaces for sequential two-means, basic two-means, exact-spike inclusion summaries, the signal adaptive variable selector and an adaptive-L1 approximation to decoupling shrinkage and selection. Both interfaces provide native Gaussian ridge, horseshoe and exact spike-and-slab fitting, posterior prediction, chain diagnostics and plots. R also retains its bayesreg interface. Separate manuals document their inputs, priors and numerical conventions. We compare fixed summaries across languages and check fitted posteriors against independent numerical integration. Four Python inputs meet the declared moment-error and chain-diagnostic thresholds; two R fits meet separate moment-error tolerances. These bounded checks support the specified calculations, without establishing general selection accuracy or reliable mixing in high-dimensional problems.