Xiao-Li Meng

Overview

Xiao-Li Meng is a statistician at Harvard University (Department of Statistics). His research focuses on Bayesian computation, EM algorithms, multiple imputation, and model assessment. He coined the term "posterior predictive p-value" and provided the foundational theoretical analysis of their properties.

Key Contributions

• Meng (1994): "Posterior Predictive p-values." Annals of Statistics 22: 1142–1160. Proved that ppc p-values are conservative (stochastically larger than Uniform in absolute deviation from 0.5) under the true model; connected this conservatism to positive correlation between $y$ and replicated $y^{rep}$.
• Gelman-Meng-Stern (1996): Co-developed the realized discrepancy framework and unified taxonomy of Bayesian replication types. See Posterior Predictive Check.
• EM algorithm extensions and missing-data theory; multiple imputation fraction-of-missing-information results.

Related

• Andrew Gelman
• Hal Stern
• Donald Rubin
• Posterior Predictive Check