Overview
Statistician affiliated with the National Institute of Statistical Sciences (NISS) and the Institute of Statistics and Decision Sciences, Duke University, at the time of his 1995 technical report. Developed the hybrid Markov chain approach for Bayesian inference in the multinomial probit model, later published as Nobile (1998, Statistics and Computing 8: 229–242).
Key Contributions / Features
- Hybrid Markov chain for MNP (1995/1998): Identified scale non-identification of the MNP likelihood as the root cause of poor Gibbs sampler mixing in McCulloch-Rossi (1994). Proposed appending a Metropolis rescaling step after each Gibbs cycle; acceptance probability requires only prior density evaluation (likelihood ratio cancels exactly). Formally proved ergodicity of the hybrid chain.
- The published version (Nobile 1998) contains an error in the acceptance ratio later corrected by Imai-van Dyk (2005), who reinterpreted the step as marginal data augmentation.
Related