(SC587) -  Ioannis Ntzoufras

Περιγραφή Μαθήματος

This short course is intended to provide an introduction to Bayesian variable selection methods. The theoretical aspects are complemented with practical examples using MCMC methods via R and MCMC sampling software (such as OpenBUGS and JAGS). The theoretical part will introduce the notions of Bayes Factors, posterior model odds and posterior model probabilities. Focus will be given in Objective Bayes model comparisons with detailed description to the popular prior formulations (such as the g-prior and the hyper-g prior) and the criteria, which ensure a well implemented variable selection method.  The theory will conclude with a presentation of the Bayesian version of LASSO methods. The practical approach will describe the conjugate case, MCMC methods for the model space (using R and associated packages) in normal linear regression. Also the implementation in non-conjugate case using variable selection MCMC methods will be also presented using WinBUGS.

The course is intended for statisticians and practitioners who wish to understand and implement modern Bayesian variable selection techniques to practical problems. Participants are advised to bring their own laptops for the lab sessions of the course.

Ημερομηνία δημιουργίας

Παρασκευή 20 Σεπτεμβρίου 2019