An R-package for Bayesian variable selection, model choice, and regularized estimation for (spatial) generalized additive mixed regression models via stochastic search variable selection with spike-and-slab priors.
- Fits additive models for Gaussian, Binary/Binomial and Poisson responses
- (Correlated) random effects
- Automagically performs variable selection and function selection (i.e., do I need this model term at all, is it linear or is it non-linear?), also for interactions between multiple covariates.
- Yields marginal posterior inclusion probabilities for each term as well as posterior model probabilities and model-averaged effect estimates.
- Convenient formula-based model specification
- Fully Bayesian via MCMC, multiple parallelized chains for diagnostics and faster mixing, sampler implemented in
C.
Short & applied intro (also the vignette that comes with the package, with some minor modifications...):
Fabian Scheipl. (2011)
spikeSlabGAM: Bayesian Variable Selection, Model Choice and Regularization for Generalized Additive Mixed Models in R. Journal of Statistical Software, 43(14). [pdf]
More theory, simulation studies and real-world case studies:
Fabian Scheipl, Ludwig Fahrmeir, Thomas Kneib (2012). Spike-and-Slab Priors for Function Selection in Structured Additive Regression Models. Journal of the American Statistical Association, 107(500), 1518-1532. [pdf on arXiv]