时间:2018年4月25(星期三)13:30-14:30
地点:学院南路校区,学术会堂603
报告题目:BayesianInference for the One-Factor Copula Model
报告人:Anastasios Panagiotelis, Monash University
摘要:We develope efficient Bayesian inference for the one-factor copula model with two significantcontributions over existing methodologies. First, our approach leads to straightforwardinference on dependence parameters and the latent factor; only inference on theformer is available under frequentist alternatives. Second, we develop areversible jump Markov chain Monte Carlo algorithm that averages over modelsconstructed from different bivariate copula building blocks. Our approachaccommodates any combination of discrete and continuous margins. Throughextensive simulations, we compare the computational and Monte Carlo efficiencyof alternative proposed sampling schemes. The preferred algorithm providesreliable inference on parameters, the latent factor and model space. Thepotential of the methodology is highlighted in an empirical study of ten binarymeasures of socio-economic deprivation collected for 11463 East Timoresehouseholds. The importance of conducting inference on the latent factor ismotivated by constructing a poverty index using estimates of the factor. Comparedto a linear Gaussian factor model, our model average improves out-of-samplefit. The relationships between the poverty index and observed variablesuncovered by our approach are diverse and allow for a richer and more preciseunderstanding of the dependence between overall deprivation and individualmeasures of well-being.
报告人简介:Anastasios Panagiotelis 是多元分析和统计计算领域的权威专家,在本领域国际著名期刊Journal ofthe American Statistical Association、Journal of Econometrics等发表发表学术论文多篇,同时还主持了多项澳大利亚国家科学基金项目。此外还担任Journal ofEconometrics, Journal of Economic Dynamics and Control, International Journalof Forecasting, Computational Statistics and Data Analysis, Technometrics,Australian and New Zealand Journal of Statistics, Electronic Journal ofStatistics, and the Journal of Statistical Software, Competitive Grant for theEstonian Research Council等多个国际统计期刊的审稿。2016年Panagiotelis教授作为中财讲席教授来威尼斯wns885566工作两个月,本次PanagiotelisPanagiotelis访问受2018威尼斯wns885566引智项目的团队式式项目支持。