Fits a Bayesian group-sparse multi-task regression model using Gibbs sampling. The hierarchical prior encourages shrinkage of the estimated regression coefficients at both the gene and SNP level. The model has been extended to a spatial model that allows for two type correlation in neuroimaging genetics data and been applied successfully to imaging phenotypes of dimension up to 100; it can be used more generally for multivariate (non-imaging) phenotypes.
| Version: | 0.5 |
| Depends: | R (≥ 3.3.0), Matrix (≥ 1.2.6), mvtnorm (≥ 1.0.5), matrixcalc (≥ 1.0.3), miscTools (≥ 0.6.22) |
| Imports: | coda (≥ 0.18.1), EDISON (≥ 1.1.1), statmod (≥ 1.4.26), methods (≥ 3.3.3), sparseMVN (≥ 0.2.0), inline (≥ 0.3.15), LaplacesDemon (≥ 16.1.0), CholWishart (≥ 0.9.3), mnormt (≥ 1.5.4), Rcpp (≥ 0.12.14) |
| Published: | 2019-01-07 |
| Author: | Yin Song, Shufei Ge, Liangliang Wang, Farouk S. Nathoo, Keelin Greenlaw, Mary Lesperance |
| Maintainer: | Yin Song <yinsong at uvic.ca> |
| License: | GPL-2 |
| NeedsCompilation: | no |
| CRAN checks: | bgsmtr results |
| Reference manual: | bgsmtr.pdf |
| Package source: | bgsmtr_0.5.tar.gz |
| Windows binaries: | r-devel: bgsmtr_0.5.zip, r-release: bgsmtr_0.5.zip, r-oldrel: bgsmtr_0.5.zip |
| OS X binaries: | r-release: bgsmtr_0.5.tgz, r-oldrel: bgsmtr_0.1.tgz |
| Old sources: | bgsmtr archive |
Please use the canonical form https://CRAN.R-project.org/package=bgsmtr to link to this page.