glmsusie

---

The glmsusie package implements the generalized Sum of Single Effects (gSuSiE) model, a likelihood-based extension of the Bayesian SuSiE framework for variable selection in generalized linear and survival (Cox) regression models. gSuSiE is designed for settings with highly correlated predictors and sparse true signals, where traditional penalized regression methods often struggle.

At its core, gSuSiE decomposes regression effects into additive “single-effect” components, enabling stable variable selection and interpretable uncertainty quantification through credible sets (CSs) — groups of correlated variables that collectively have high probability of containing at least one true effect.

Beyond credible sets and posterior inclusion probabilities (PIPs), gSuSiE also provides standard errors (SEs), p-values, and 2BIC values as additional diagnostic measures, offering both Bayesian and frequentist perspectives on variable importance.

Developed by Yizeng Li and Wei Pan, Division of Biostatistics and Health data Science, University of Minnesota Twin Cities.

Quick start

You can install the latest development version of glmsusie below:

# install.packages("remotes")
remotes::install_github("yizenglistat/glmsusie")

See here for a brief illustraton of glmsusie. For more documentation and examples please visit https://github.com/yizenglistat/glmsusie.