Parameter Expansion in Bayesian Hierarchical Modeling

This is the project that I did in my Bayeisan Statistics course, below are the snapshots of my part of my paper.

Abstract: Hierarchical model is devoted to facilitate the simultaneous estimation of severalparameters over similar units. However, some problems pertinent to Bayesian hierarchicalmodeling remain unsolved, that is: if the standard deviation for the second layer of hierarchicalmodel (also called between-study standard deviation) has broad peak at zero, somenoninformative prior such as flat uniform prior, IG(0.01,0.01) which are normally adopted inresearch, may lead to insensitivity in the estimation of such smoothing variance. Apart fromthis, in this prior setting, convergence based on EM and Gibbs sampling may become lessconvergence. In this paper, we bring forward a multiplicative parameter-expansion methodto reparameterize hierarchical model in the context of Bayesian inference which facilitatesconvergence and possesses decent properties. Illustrations in terms of simulation will bedelivered to reveal the two-fold essences of this method.

Require for paper"Parameter Expansion in Bayeisan Hierarchical Modeling" ? Simply send an email to stefanie.cao@gmail.com with title "require paper hierarchical modeling".