基于小波方法的时变长记忆参数的估计.pdf

《基于小波方法的时变长记忆参数的估计.pdf》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、应用概率统计第二十八卷ChineseJournalofAppliedProbability第五期2012年10月andStatisticsVo1．28No．5Oct．2012Time．VaryingLongMemoryParameterEstimationBasedoneletsLUZHIPING(ResearchCenterofInternationalFinancialandRiskManagement；SchoolofFinanceandStatistics，EastChinaNormalUniversity

2、，Shanghai，200241)TAOQINYINGfDeparmentofMathem。tics，EastChin。Norm。lUniversity，Shanghni，200241)AbstractStationarvlongmemoryprocesshasbeenwidelystudiedintheliterature．Inthisarticle，weconsideredthelocallystationary1ongmemoryprocesswithtimevaryingmemoryparameter·An

3、ewwavelet—basedalgorithmWaSdevelopedusinglog—linearrelationshipbetweenthewaveletcoeficientvarianceandthescalingparameter．Theconsistencyandthefinitesamplebehavioroftheestimatorhrdvealsobeenstudied，whichprovideagoodrefereneeforthepractitionerandresearchers．Thene

4、walgorithmhasalsobeenappliedtotheYEN／USDexchangerateseries，whichleadstosomeinterestingresults．Keywords"Locallystationarylongmemoryprocess，semi—parametricestimation，waveletanalysis，exchangerate．AMSSubjectC：lassification：62M09．62M10．62—07．§1．IntroductionLongmemo

5、rvmodelhasreceivedconsiderableattentionssinceitsfirstproposalbyGrangerandJoyeux(1980)andHosking(1981)．Ithasbeenappliedinavarietyoffields，forexample，hydrology,geography，engineering，finance，economy，climatology，medicine，biology,etc．．Themostwidelyusedstationarylon

6、gmemorymodelistheFractionallyIntegratedAutoRegressiveMovingAverage(FARIMA)model，definedbythefollowingequation：f—B)dXt=￡t，where￡￡isanARMA(p，q)process，d∈(0，0．5)andBisthebackshiftoperator．Insteadoffirst—orderdiference，theseriesisd-thdiferenced．Thisworkwassupporte

7、dby“theFundamentalResearchFundsfortheCentralUniversities’’andtheNationalNaturalScienceFoundationofChina(11101158)．Correspondingauthor，E—mail：taoqinying@hotmail．com．ReceivedNovember1，2011．RevisedMay4，2012．500应用概率统计第二十八卷Intuitively，treatingthelongmemoryparameter

8、asaconstantimpliesthatthelongrangedependencestructureoftheunderlyingphenomenonpersistsovertime．However，sometimestheassumptionofconstantlongmemoryparameteristoorestrictive．Inprac—ti