A comparison of direct and iterated multistep AR methods .pdf

A comparison of direct and iterated multistep AR methods .pdf

ID:49865441

大小:181.56 KB

页数:31页

时间:2020-03-05

A comparison of direct and iterated multistep AR methods .pdf_第1页
A comparison of direct and iterated multistep AR methods .pdf_第2页
A comparison of direct and iterated multistep AR methods .pdf_第3页
A comparison of direct and iterated multistep AR methods .pdf_第4页
A comparison of direct and iterated multistep AR methods .pdf_第5页
资源描述:

《A comparison of direct and iterated multistep AR methods .pdf》由会员上传分享,免费在线阅读,更多相关内容在应用文档-天天文库

1、AComparisonofDirectandIteratedMultistepARMethodsforForecastingMacroeconomicTimeSeriesMassimilianoMarcellinoIstitutodiEconomiaPolitica,UniversitaBocconiandIGIERJamesH.StockDepartmentofEconomics,HarvardUniversityandtheNBERandMarkW.Watson*DepartmentofEconomicsandWoodrowWil

2、sonSchool,PrincetonUniversityandtheNBERFebruary2004(RevisedApril2005)*TheauthorsthankJin-LungLin,FrankSchorfheide,KenWest,andtworefereesforcomments.ThisresearchwasfundedinpartbyNSFgrantSBR-0214131.Abstract“Iterated”multiperiodaheadtimeseriesforecastsaremadeusingaone-per

3、iodaheadmodel,iteratedforwardforthedesirednumberofperiods,whereas“direct”forecastsaremadeusingahorizon-specificestimatedmodel,wherethedependentvariableisthemulti-periodaheadvaluebeingforecasted.Whichapproachisbetterisanempiricalmatter:intheory,iteratedforecastsaremoreef

4、ficientiftheone-periodaheadmodeliscorrectlyspecified,butdirectforecastsaremorerobusttomodelmisspecification.Thispapercomparesempiricaliteratedanddirectforecastsfromlinearunivariateandbivariatemodelsbyapplyingsimulatedout-of-samplemethodsto170U.S.monthlymacroeconomictime

5、seriesspanning1959–2002.Theiteratedforecaststypicallyoutperformthedirectforecasts,particularlyifthemodelscanselectlonglagspecifications.Therelativeperformanceoftheiteratedforecastsimproveswiththeforecasthorizon.Keywords:multistepforecasts,VARforecasts,forecastcomparison

6、sJELNumbers:C32,E37,E471.IntroductionAforecastermakingamultiperiodtimeseriesforecast–forexample,forecastingtheunemploymentratesixmonthshence–confrontsachoicebetweenusingaone-periodmodeliteratedforward,orinsteadusingamultiperiodmodelestimatedwithalossfunctiontailoredtoth

7、eforecasthorizon.Inthecaseofunivariatelinearmodelsandquadraticloss,the“iterated”forecast(sometimescalleda“plug-in”forecast)entailsfirstestimatinganautoregression,theniteratinguponthatautoregressiontoobtainthemultiperiodforecast.Incontrast,theforecastbasedonthemultiperio

8、dmodel–which,followingtheliterature,weshallcallthe“direct”forecast–entailsregressingamultiperiod-aheadvalueoft

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。