LOCUS – Large-scale variational inference for Bayesian variable selection in multiple-response regression 
locus is an R package providing efficient variational algorithms for simultaneous variable selection of covariates and associated responses based on multivariate regression models. Dependence across responses linked to the same covariates is captured through the model hierarchical structure (H. Ruffieux, A. C. Davison, J. Hager, I. Irincheeva, Efficient inference for genetic association studies with multiple outcomes, Biostatistics, 18:618–636, 2017).
To install, run the following command in R:
if(!require(remotes)) install.packages("remotes")
remotes::install_github("hruffieux/locus")The algorithms for joint covariate and response selection provided in locus implement inference for regression models with
- identity link;
- logistic link;
- probit link;
- identity-probit link.
Inference on models for group selection and based on a spatial Gaussian process to encode the dependence structure of the candidate predictors are also implemented. Moreover, covariate-level external information variables can be incorporated to inform the selection.
This software uses the GPL v2 license, see LICENSE. Authors and copyright are provided in DESCRIPTION. Loris Michel has also contributed to the development of this project.
Please cite the software using the following reference: H. Ruffieux, A. C. Davison, J. Hager, I. Irincheeva, Efficient inference for genetic association studies with multiple outcomes, Biostatistics, 18:618–636, 2017.
To report an issue, please use the locus issue tracker at github.com.