欢迎来到天天文库
浏览记录
ID:40709866
大小:3.50 MB
页数:81页
时间:2019-08-06
《Chapter-57-Markov-Chain-Monte-Carlo-Methods-Computation-and-Inference_2001_Handbook-of-Econometrics》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、Chapter57MARKOVCHAINMONTECARLOMETHODS:COMPUTATIONANDINFERENCESIDDHARTHACHIB*JohnMOlinSchoolofBusiness,WashingtonUniversity,CampusBox1133,1BrookingsDr,StLouis,MO63130,USAContentsAbstract3570Keywords35701Introduction35711.1Organization35732Classicalsamplingmethods3573
2、2.1Inversetransformmethod35732.2Accept-rejectalgorithm35752.3Methodofcomposition35763Markovchains35763.1Definitionsandresults35773.2Computationofnumericalaccuracyandinefficiencyfactor35794Metropolis-Hastingsalgorithm35804.1Thealgorithm35814.2Convergenceresults35844.
3、3Example35854.4Multiple-blockM-Halgorithm35875TheGibbssamplingalgorithm35895.1Thealgorithm35905.2Connectionwiththemultiple-blockM-Halgorithm35915.3InvarianceoftheGibbsMarkovchain35925.4Sufficientconditionsforconvergence35925.5Estimationofdensityordinates35925.6Examp
4、le:simulatingatruncatedmultivariatenormal35946Samplerperformanceanddiagnostics35957Strategiesforimprovingmixing35967.1Choiceofblocking3597'email:chib@olinwustleduHandbookofEconometrics,Volume5,EditedbyJJHeckmanandELeamer©2001ElsevierScienceBVAllrightsreserved3570SCh
5、ib7.2Tuningtheproposaldensity35977.3Otherstrategies35988MCMCalgorithmsinBayesianestimation35998.1Overview35998.2Notationandassumptions36008.3Normalandstudent-tregressionmodels36028.4Binaryandordinalprobit36048.5Tobitcensoredregression36078.6Regressionwithchangepoint
6、36088.7Autoregressivetimeseries36108.8HiddenMarkovmodels36128.9Statespacemodels36148.10Stochasticvolatilitymodel36168.11Gaussianpaneldatamodels36198.12Multivariatebinarydatamodels36209Samplingthepredictivedensity362310MCMCmethodsinmodelchoiceproblems362610.1Backgrou
7、nd362610.2Marginallikelihoodcomputation362710.3Modelspace-parameterspaceMCMCalgorithms363310.4Variableselection363610.5Remark363911MCMCmethodsinoptimizationproblems363912Concludingremarks3641References3642AbstractThischapterreviewstherecentdevelopmentsinMarkovchainM
8、onteCarlosim-ulationmethodsThesemethods,whichareconcernedwiththesimulationofhighdimensionalprobabilitydistributions,havegainedenormousprom
此文档下载收益归作者所有
点击更多查看相关文章~~