Advances and Applications in Statistics
Volume 7, issue 2, Pages 291 - 302
(August 2007)
|
|
NORM COMPARISONS FOR DATA AUGMENTATION
James P. Hobert (U. S. A.) and Jeffrey S. Rosenthal (Canada)
|
Abstract: We consider the convergence and efficiency of various data augmentation algorithms, including the parameter-expansion data augmentation (PX-DA) algorithms of Liu and Wu [8], Meng and van Dyk [9], and Hobert and Marchev [5]. In particular, we explore connections between Markov chain partial order introduced by Peskun [12], operator norm bounds, geometric ergodicity, variance bounding Markov chains, and L2 theory. Our main result is a direct generalisation of one of the theorems in Hobert and Marchev [5]. |
Keywords and phrases: Markov chain, MCMC, data augmentation, PX-DA, Peskun order, geometric ergodicity, variance bounding. |
|
Number of Downloads: 340 | Number of Views: 1045 |
|