Overview
David A. van Dyk is a professor in the Department of Statistics at the University of California, Irvine (at the time of the paper). He works on Bayesian computation, data augmentation, and MCMC methods. He is co-developer (with Xiao-Li Meng) of the conditional and marginal data augmentation framework (Meng-van Dyk 1999; van Dyk-Meng 2001), and co-developer (with Kosuke Imai) of the marginal augmentation algorithm for the multinomial probit model.
Key Contributions
- Meng and van Dyk (1999): "Seeking Efficient Data Augmentation Schemes via Conditional and Marginal Augmentation" — introduces the marginal augmentation framework; proves that averaging over a working prior distribution for an unidentifiable parameter can only improve the geometric rate of convergence of the DA algorithm. See Data Augmentation.
- van Dyk and Meng (2001): "The Art of Data Augmentation" — comprehensive treatment of the conditional/marginal augmentation framework.
- Imai and van Dyk (2005): Applies marginal augmentation to the multinomial probit model; new prior on identifiable parameters; corrects Nobile (1998) error; Algorithm 1 Scheme 1 dominates all existing methods. See Multinomial Probit and Imai-van Dyk (2005).
Related