Overview
L. Mark Berliner is a statistician at Ohio State University. He co-authored with Steven N. MacEachern the definitive proof that thinning a stationary Markov chain output is always variance-inefficient.
Key Contributions
- Co-authored MacEachern-Berliner (1994), proving that systematic subsampling of a Gibbs sampler never improves posterior mean estimates and the only legitimate reason to thin is computational.
Related