A header only R package collects multivariate Spike-and-Slab model with effective posterior explorations by Shen, Solis-Lemus and Deshpande 2022.
Method implemented:
-
For continuous data
- multivariate SSL by Deshpande et al. 2018, this uses a mean covariance parameterization of Gaussian model
- chain graphical SSL by Shen, Solis-Lemus and Deshpande 2022, uses a Gaussian chain graph parameterization
- chain graphical VAR(1) SSL described in the same Shen, Solis-Lemus and Deshpande 2021+, uses a Gaussian chain graphical VAR(1) model for time series
-
For binary data
- multivariate probit SSL follows multivariate Spike-and-Slab LASSO by Deshpande et al. 2018, with an EM algorithm account for the probit structure follows Gessner et al. 2019's sampling methods.
- the chain graphical version of the above
-
For counting
- multivariate STAR SSL (currently only allow fixed links) follows multivariate Spike-and-Slab LASSO by Deshpande et al. 2018, with EM algorithm described in Kowal and Canale 2020's STAR model.
- cg version of the above
It is not yet on CRAN to install, use:
devtools::install_github("YunyiShen/mSSL")For experiments and real data analysis in our paper, please refer to the dev branch for more information.