Edward Greenberg is an economist at Washington University in St. Louis (Department of Economics). He co-authored two foundational 1995 papers with Siddhartha Chib: the canonical MH tutorial (Chib-Greenberg 1995a) and a paper developing hierarchical Bayesian inference for SUR models with correlated errors and time-varying parameters via MCMC (Chib-Greenberg 1995b). He also co-authored with Robert P. Parks a predictive Bayesian approach to model selection (Greenberg-Parks 1997).
Chib-Greenberg (1998) multivariate probit — with Siddhartha Chib; unified simulation-based Bayesian and MLE framework for correlated binary data via the multivariate probit model; three-block Gibbs (latent data via Geweke 1991 multivariate truncated Normal, conjugate Normal for β, tailored independence/reflection M-H for correlation matrix σ with Newton-Raphson mode and inverse-Hessian proposal); extends Chib (1995) marginal likelihood identity to unknown-normalising-constant settings via kernel density estimation; GHK likelihood (eq. 11); MCEM for MLE via Louis (1982) information; three applications: voter behaviour (), Six Cities wheezing (, equi-correlated model wins), PSID labour force (, 21 free correlations, unrestricted model wins decisively). See Multinomial Probit and Marginal Data Density.
Chib-Greenberg (1995a): With Siddhartha Chib; tutorial derivation of MH from reversibility, unification of five candidate-generating families, Product of Kernels principle establishing Gibbs as special case of MH, M-H A-R algorithm (§6.1), optimal acceptance rate guidance from Roberts-Gelman-Gilks (1994). See Chib-Greenberg (1995).
Chib-Greenberg (1995b): With Siddhartha Chib; Gibbs sampling for hierarchical SUR; Metropolis-within-Gibbs for VMA(1) errors (Taylor-approximation candidate at NLS estimate, ~50% acceptance); joint FFBS backward simulation for TVP-SUR states; partial Bayes factors for model comparison; OECD GNP application confirms pooled model (log PBF ≈ 40) and rejects time-varying parameters. See Chib-Greenberg (1995b).
Greenberg-Parks (1997): With Robert P. Parks; predictive Bayesian model selection via comparison of Student-t predictive densities; GVR (generalized variance ratio) and overlap statistics; Bayesian interpretation of multicollinearity; application to the Fazzari-Hubbard-Petersen investment model showing cash flow dominates Tobin's q in prediction.