資源描述:
《時(shí)間序列的經(jīng)驗(yàn)似然擬合優(yōu)度檢驗(yàn)》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在行業(yè)資料-天天文庫(kù)。
1、目錄Contents中文摘要???????????????????1英文摘要???????????????????2時(shí)間序列的經(jīng)驗(yàn)似然擬合優(yōu)度檢驗(yàn)????????????31引論???????????????????.1.1背景介紹????????..?????????..1.2本文主要內(nèi)容和創(chuàng)新點(diǎn)??????????.????2局部多項(xiàng)式估計(jì)????????????????.63經(jīng)驗(yàn)似然比?????????????????.93.1經(jīng)驗(yàn)似然的建立??????????.??????93.2A(z)的估計(jì)??????..?.?.????.????9
2、3.3似然比的估計(jì)?.?????.??????.????154擬合優(yōu)度檢驗(yàn)?????????????????175總結(jié)與展望?????????????????.195.1內(nèi)容總結(jié)............?.?...?.??.?....?????19IJMIIJIJIIIJlIIHIPIflllllrJllIY23729125.2未來(lái)展望??????????????????.19參考文獻(xiàn)???????????????????20致謝????????????????????.24中文摘要畢業(yè)論文題目:時(shí)間序列的經(jīng)驗(yàn)似然擬合優(yōu)度檢墜概率論與數(shù)理統(tǒng)計(jì)專業(yè)
3、—20—09級(jí)碩士生姓名:董望指導(dǎo)教師(姓名、職稱):至皇迭塾蕉時(shí)間序列的擬合優(yōu)度檢驗(yàn)是統(tǒng)計(jì)理論中非常重要的內(nèi)容。在獨(dú)立同分布的情形下,前人已經(jīng)做了較多研究。對(duì)于相依數(shù)據(jù)的研究還有待完善。經(jīng)驗(yàn)似然是Owen(1988)在完全樣本下提出的一種非參數(shù)統(tǒng)計(jì)推斷方法,它有類似于bootstr印的抽樣特性。這一方法與經(jīng)典的統(tǒng)計(jì)方法比較有很多突出的優(yōu)點(diǎn),如:Wilk’S性質(zhì)和Barlett糾偏性。Chcn,HMdleandLi(2003)在前人基礎(chǔ)上充分利用經(jīng)驗(yàn)似然的優(yōu)點(diǎn),改進(jìn)了擬合優(yōu)度的檢驗(yàn)結(jié)果。本文對(duì)Chen,H心dleandLi(2003)的研究進(jìn)行了
4、進(jìn)一步地改進(jìn)。由于局部多項(xiàng)式方法在估計(jì)條件均值函數(shù)m(z)=E(Y]X=z)方面,相比于Nw核估計(jì)方法有許多更優(yōu)良的性質(zhì),所以,本文中我們引入了局部多項(xiàng)式估計(jì)代替上文中的NW估計(jì)方法來(lái)估計(jì)條件均值函數(shù),繼而建立估計(jì)量的經(jīng)驗(yàn)似然,并證明了一個(gè)引理,從而獲得經(jīng)驗(yàn)似然比的估計(jì)。然后在此基礎(chǔ)上構(gòu)造檢驗(yàn)統(tǒng)計(jì)量,并與HM統(tǒng)計(jì)量做了比較,展現(xiàn)經(jīng)驗(yàn)似然統(tǒng)計(jì)量的優(yōu)勢(shì)。最后分別用理論上的漸近分布和bootstrap兩種方法研究了該統(tǒng)計(jì)量的分布,得出了拒絕域,從而可以期望獲得更好的檢驗(yàn)結(jié)果。關(guān)鍵詞:&一混合,經(jīng)驗(yàn)似然,擬合優(yōu)度檢驗(yàn),核估計(jì),局部多項(xiàng)式估計(jì),參數(shù)模型,檢
5、驗(yàn)功效,bootstrap方法1英文摘要THESIS:AnEmpiricalLikelihoodGoodness—of-FitTestforTimeSeriesSPECIALIZATION:ProbabilityandMathematicalStatisticsPOSTGRADUATE:HuangYiMENTOR:ProfessorWangLihongThegoodness—of-fittestfortimeseriesisvitaltostatistics。Theproblemoftestingaparametricmeanregression
6、againstanonparametricalternativeisnotnewforanindependentandidenticallydistributedsetting.Yetfordependentsettingmoreeffortsareneededtocompleterelevanttheory.Owen(1998)producedEmpiricalLikelihoodasanonparametricinferencemethodforcompletesamples.Ithassamplepropertylikebootstrap.
7、Comparedtoclassicalstatistics,thismethodhasmanyprominentad—vantagessuchasWilk’SpropertyandBarlettcorrectability.Basedonthepredecessor,Chen,H打dleandLi(2003)madefulluseoftheadvantagesofempiricallikelihoodandimprovedtheresolutionofthegoodness—of-fittest.Duetotheprominentproperti
8、esofthelocalpolynomialmethodforestimatingtheconditionalmeanfunctionm