R package horseshoe: Implementation of the Horseshoe Prior. Contains functions for applying the horseshoe prior to high- dimensional linear regression, yielding the posterior mean and credible intervals, amongst other things. The key parameter tau can be equipped with a prior or estimated via maximum marginal likelihood estimation (MMLE). The main function, horseshoe, is for linear regression. In addition, there are functions specifically for the sparse normal means problem, allowing for faster computation of for example the posterior mean and posterior variance. Finally, there is a function available to perform variable selection, using either a form of thresholding, or credible intervals.
Keywords for this software
References in zbMATH (referenced in 6 articles )
Showing results 1 to 6 of 6.
- Jason Willwerscheid, Matthew Stephens: ebnm: An R Package for Solving the Empirical Bayes Normal Means Problem Using a Variety of Prior Families (2021) arXiv
- Martin, Ryan; Ning, Bo: Empirical priors and coverage of posterior credible sets in a sparse normal mean model (2020)
- Martin, Ryan; Tang, Yiqi: Empirical priors for prediction in sparse high-dimensional linear regression (2020)
- Angela Bitto-Nemling, Annalisa Cadonna, Sylvia Frühwirth-Schnatter, Peter Knaus: Shrinkage in the Time-Varying Parameter Model Framework Using the R Package shrinkTVP (2019) arXiv
- Bhadra, Anindya; Datta, Jyotishka; Polson, Nicholas G.; Willard, Brandon: Lasso meets horseshoe: a survey (2019)
- Daniel F. Schmidt, Enes Makalic: High-Dimensional Bayesian Regularised Regression with the BayesReg Package (2016) arXiv