欢迎来到天天文库
浏览记录
ID:7286095
大小:1.42 MB
页数:33页
时间:2018-02-10
《bayesian analysis utility prior and bayesian robustness》由会员上传分享,免费在线阅读,更多相关内容在工程资料-天天文库。
1、Utility,Prior,andBayesianRobustnessWebeginthischapterwithadiscussionofrationahtyaxiomsforpreferenceandhowonemaydeducetheexistenceofautihtyandprior.LaterweexplorehowrobustorsensitiveisBayesiananalysistothechoiceofprior,utility,andmodel.Intheprocess,weintroducean
2、dexaminevariousquantitativeevaluationsofrobustness.3.1Utility,Prior,andRationalPreferenceWehaveintroducedinChapter2problemsofestimationandtestingasBayesiandecisionproblems.WerecallthecomponentsofageneralBayesiandecisionproblem.LetXbethesamplespace,0theparameter
3、space,f{x )thedensityofXand7r(^)priorprobabilitydensity.Moreover,thereisaspaceAofactions"a"andalossfunctionL{0,a).Thedecisionmaker(DM)chooses"a"tominimizetheposteriorrisk^{ax)=/L{e,a)iT{ex)dO,(3.1)whereT:{9X)istheposteriordensityof0givenx.Notethatgiventhelo
4、ssfunctionandtheprior,thereisanaturalpreferenceorderingai:5、x).ThereisalongtraditionoffoundationalstudydatingbacktoRamsey(1926),inwhichonestartswithsuchapreferencerelationon.4x^satis•fyingcertainrationalaxioms(i.e6、.,axiomsmodelingrationalbehavior)liketransitivity.Itcanthenbeshownthatsucharelationcanonlybeinducedasaboveviaalossfunctionandaprior,i.e.,3LandTTsuchthatai7、on,onecanrecoverthesubjectivelossfunctionandprior.Ifthereisnosampledata.663Utility,Prior,andBayesianRobustnessthenTVwouldqualifyasasubjectivepriorfortheDM.Ifwehavedatax,alikelihoodfunctionf{x6)isgivenandweareexaminingapreferencerelationgivenx,thenalsoonecanded8、ucetheexistenceofLandTTsuchthatai:9、musthavea(subjective)priorifourpreferenceamongactionssatisfiescertainaxiomsaboutrationalbehavior.Together,theyjustify(3.2)andthrowsomelighton(3.3).Intheremainingsectionsweex
5、x).ThereisalongtraditionoffoundationalstudydatingbacktoRamsey(1926),inwhichonestartswithsuchapreferencerelationon.4x^satis•fyingcertainrationalaxioms(i.e
6、.,axiomsmodelingrationalbehavior)liketransitivity.Itcanthenbeshownthatsucharelationcanonlybeinducedasaboveviaalossfunctionandaprior,i.e.,3LandTTsuchthatai7、on,onecanrecoverthesubjectivelossfunctionandprior.Ifthereisnosampledata.663Utility,Prior,andBayesianRobustnessthenTVwouldqualifyasasubjectivepriorfortheDM.Ifwehavedatax,alikelihoodfunctionf{x6)isgivenandweareexaminingapreferencerelationgivenx,thenalsoonecanded8、ucetheexistenceofLandTTsuchthatai:9、musthavea(subjective)priorifourpreferenceamongactionssatisfiescertainaxiomsaboutrationalbehavior.Together,theyjustify(3.2)andthrowsomelighton(3.3).Intheremainingsectionsweex
7、on,onecanrecoverthesubjectivelossfunctionandprior.Ifthereisnosampledata.663Utility,Prior,andBayesianRobustnessthenTVwouldqualifyasasubjectivepriorfortheDM.Ifwehavedatax,alikelihoodfunctionf{x6)isgivenandweareexaminingapreferencerelationgivenx,thenalsoonecanded
8、ucetheexistenceofLandTTsuchthatai:9、musthavea(subjective)priorifourpreferenceamongactionssatisfiescertainaxiomsaboutrationalbehavior.Together,theyjustify(3.2)andthrowsomelighton(3.3).Intheremainingsectionsweex
9、musthavea(subjective)priorifourpreferenceamongactionssatisfiescertainaxiomsaboutrationalbehavior.Together,theyjustify(3.2)andthrowsomelighton(3.3).Intheremainingsectionsweex
此文档下载收益归作者所有