A.F.M. Smith

Overview

A.F.M. (Adrian Frederick Melhuish) Smith is a Bayesian statistician, formerly at Imperial College London (Department of Mathematics) and later at Queen Mary University of London and the University of Warwick, and subsequently Director of the Alan Turing Institute. He is a central figure in the development of Markov Chain Monte Carlo methods, including co-authoring the foundational Gelfand-Smith (1990) paper that introduced Gibbs sampling to statistics and the George-Makov-Smith (1993) paper on conjugate likelihood distributions.

Key Contributions

• Gelfand-Smith (1990): Co-introduced Gibbs sampling as a general-purpose Bayesian computation tool; "Sampling-Based Approaches to Calculating Marginal Densities" established the Griddy-Gibbs algorithm and launched the MCMC revolution in Bayesian statistics.
• George-Makov-Smith (1993): Co-developed conjugate likelihood distributions for exponential-family hierarchical models; Theorem 3.3 establishes log-concavity and the $p\geq 2$ properness condition enabling exact Gibbs sampling via Gilks-Wild ARS.

Related

• Bayesian Hierarchical Model
• Gibbs Sampler
• Edward I. George
• Udi E. Makov
• George-Makov-Smith (1993)