Publications [#50249] of Mark Huber

Papers Submitted

1. F. Mitha and M. Huber, Monotonic Multigamma coupling for perfect sampling, TOMACS (2006)
   (last updated on 2006/11/10)

   Author's Comments:
   Based on Faheem Mitha's thesis at UNC. I was a committee member for this thesis.

   Abstract:
   We employ a monotonic version of the Multgamma Coupler introduced by Murdoch and Green (1998) in the context of bounded multivariate distributions, using the Gibbs sampler. Use of monotonicity greatly increases the fficiency of the coupler. In addition we observe that the multgamma coupler does not require the value of the rho parameter to be constant, further improving efficiency. We apply our algorithm on a pump dataset also examined by Murdoch and Green in order to compare the respective algorithms.