题目：Improved Marginal Likelihood Estimation via Power Posteriors and Importance Sampling

主讲人：新加坡管理大学  YU JUN教授

主持人：统计学院 常晋源教授

时间：2020年9月18日（周五）上午10:00-11:20

直播平台及会议ID：腾讯会议，257 567 496

报告摘要：

Power posteriors have become popular in estimating the marginal likelihood of a Bayesian model. Important power-posterior-based algorithms include thermodynamic integration (TI) of Friel & Pettitt (2008) and steppingstone sampling (SS) of Xie et al. (2011). In this paper, it is shown that the Bernstein-von Mises (BvM) theorem holds for power posteriors under regularity conditions. Due to the BvM theorem, power posteriors, when adjusted by the square root of the auxiliary constant, has the same limit distribution as the original posterior distribution, facilitating the implementation of the improved TI and SS methods via importance sampling. Unlike the TI and SS methods that require repeated sampling from the power posteriors, the improved methods only require the original posterior output and hence, are computationally more efficient. Moreover, they completely avoid the coding efforts associated with sampling from the power posteriors. Primitive conditions, under which the TI and improved TI algorithms can produce consistent estimators of the marginal likelihood, are provided. The numerical efficiency of the proposed methods is illustrated using two models.

主讲人简介：

YU JUN教授是新加坡管理大学李光前经济学与金融学教授 , Journal of Econometrics会士，金融计量经济学会（SoFiE）会士、理事会委员。研究领域为金融计量经济学、资产定价以及计量经济学理论。担任国际权威学术期刊Journal of Econometrics、Econometric Theory、Journal of Financial Econometrics副主编，2019年至今担任Journal of Econometrics“经济学与金融学中的贝叶斯方法”特刊特邀主编。截至2020年9月，科研成果谷歌引用次数为6876次，H指数为33, H10指数为68。2014年以首席专家身份主持新加坡国家重点课题（AcRF Tier-3）“人口老龄化的经济学挑战”，获当时新加坡历史上对社会科学研究项目授予的最高研究经费。曾担任国际货币基金组织技术顾问，为东盟+3（ASEAN+3）宏观经济研究办公室、新加坡金融管理局、新加坡政府投资公司做过咨询。