Bayesian Computation Without Tears

• Activity Number: 172
• Type: Topic Contributed
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Computing
• Author(s): Mani Lakshminarayanan*+
• Companies: Merck Research Laboratories
• Keywords: Posterior distribution ; Predictive Probability ; Markov Chain Monte Carlo ; Bayesian software ; Bayesian modeling

Abstract

Interest in the application of Bayesian methods in various scientific fields has been increasing constantly over the past decade or so. One of the primary reasons for such a growth is the development and implementation of computing algorithms such as Markov Chain Monte Carlo (MCMC) methods in 1990s that have made Bayesian computations more tractable in more complex models that could not be handled before this innovation. With the introduction of BUGS (Bayesian Inference Using Gibbs Sampling) language for specifying complex Bayesian models, followed by its Windows version, WinBUGS, practitioners across the scientific spectrum were introduced to complex Bayesian calculations that were almost nonexistent prior to late 1980s. Besides BUGS, the last decade has seen an explosion of Bayesian tools written in R (eg, BRUGS, R2WinBUGS and others) and SAS that have expanded the options for Bayesian computing to a broader audience. In this presentation, our focus will be on introducing simple Bayesian computations (after introducing simple Bayesian ideas) to an audience who has never been exposed to or who are not involved in Bayesian calculations on a regular basis.