Two convergence properties of hybrid samplers (Q1296738): Difference between revisions

This note uses the result of the authors [Electron. Commun. Probab. 2, 13-25 (1997; Zbl 0890.60061)] which studied geometric ergodicity properties of hybrid chains in terms of their constituent component algorithms. At the beginning, it is presented the definition (taken from the above quoted paper): Let \(C= (P_1,\dots, P_k)\) be a collection of Markov kernels on state space. The random scan hybrid sampler for \(C\) is the sampler defined by \(P_{RS}={1\over k} (P_1+\cdots+ P_k)\). Two special cases are considered. In the first case, there appears a quantitative result for the rate of convergence of resulting hybrid algorithms although this is at the expense of imposing a very strong uniform type of geometric ergodicity of the constituent algorithms. In the second case concerning the combination of various Metropolis algorithms and adapting the result of \textit{G. O. Roberts} and \textit{R. L. Tweedie} [Biometrika 83, No. 1, 95-110 (1996; Zbl 0888.60064)] geometric ergodicity is established. For other details see the authors' references.