Abstract: In the variable selection problem, the uncertainty regarding which is the true model is explicitly considered. In this setting, the competing models differ about which subset of variables are to be included as explanatory covariates for a given dependent variable. The Bayesian solution to this problem is conceptually straightforward and any feature of interest (e.g. inclusion probabilities of covariates or posterior predictions) is a deterministic function of the posterior distribution over the model space. This talk will focus on two different aspects regarding this distribution that need special attention. First one is theoretical and is about the prior distribution to be used for the parameters within each model since it is well known that results are dramatically sensible to this input. Second aspect, of a computational nature, is how to explore the posterior distribution since the exhaustive enumeration of all the models entertained is normally not feasible and features of interest have to be approximated/estimated based on the very small proportion of models visited.