Chapter 10 Bayesian model averaging in variable selection

We outline in this chapter a framework for addressing model uncertainty and averaging across different models in a probabilistically consistent manner. The discussion tackles two major computational challenges in Bayesian model averaging: the vast space of possible models and the absence of analytical solutions for the marginal likelihood.

We begin by illustrating the approach within the Gaussian linear model, assuming exogeneity of the regressors, and extend the analysis to cases with endogenous regressors, and dynamic models. Additionally, we adapt the framework to generalized linear models, including the logit, gamma, and Poisson families. Lastly, we explore alternative methods for computing marginal likelihoods, especially when the Bayesian information criterion’s asymptotic approximation proves inadequate.

Remember that we can run our GUI typing shiny::runGitHub("besmarter/BSTApp", launch.browser=T) in the R console or any R code editor and execute it. However, users should see Chapter 5 for details.