A0214
Title: BayMDS: An R package for Bayesian multidimensional scaling and choice of dimension
Authors: Man-Suk Oh - Ewha Womans University (Korea, South) [presenting]
Abstract: Over the last two decades, there has been a great interest in Bayesian approaches to multidimensional scaling (MDS) due to their advantages over traditional MDS methods. It provides an object configuration along with estimation errors and a simple Bayesian dimension selection criterion MDSIC for optimal dimensionality. However, Bayesian MDS (BMDS) requires a complicated Markov chain Monte Carlo (MCMC) method that may prohibit the wide use of BMDS by practitioners. A set of R functions to perform BMDS, using WinBUGS for MCMC is available. However, WinBUGS has not been updated since 2007 and it may not be efficient since it does not consider special characteristics of BMDS model. In view of these considerations, we have developed an R package bayMDS to implement BMDS that is efficient and can be easily applied by non-experts in MCMC.
