To counteract this bias, we, therefore, report credible intervals under the assumption that the effect in question is included in the model by calculating the high-density interval for each effect only across MCMC samples where the corresponding spike-and-slab variable was not 0:14\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widehat{HDI}({\beta }_{m,k})=HDI({\beta }_{m,k}|{\tau }_{m,k} \, > \, 0).$$\end{document}HDI^(βm,k)=HDI(βm,k∣τm,k>0).
