第六章-ARCH和GARCH效应的检验.docx
4页
ARCH效应的检验首先进行最小二乘法1) Q 统计量的检验：view 下 residual test 下 QstatisticDate; 04/06/11 Time; 15:11Sample: 2 1602Included obs&ivations: 1 601AutocorrelationPartial CorrelationACPACQ-StatProbI111'10.0300.0301.40050.-2.37I111■2-0.016-0.0171.82100.4 02：I11130.0170.018；^25870.520I11140.0060.0052.-31730.676I111爭-0.005-0.0052.36210.797I111-0.014-0.0142.67310.843I111'7-0.0170.0173.13570.972I111s.o.aa?0.001114360.925I1119-o.aae-0.008：3.27700.953I)1)100.0270.0274.41350.927I11111o.aae0.0054.50670.953I11112.o.ai80.0195.04590.956I11113o.aa70.0055 11690.973I111140.0210.0205.80600.971I111-15o.ai50.0136.16620.977I11116o.aae0.0096.27730.985I11117o.aae0.0076.39400.990I1111B-o.aa:-0.0036.40040.99411(11S-0.044-0.0449.51580.9641J11200.0380.04011.8000.923'[1E121-0.065-0.07018 6720.606111122o.aii0.01718.8520.65411112-3-0.027-0.03219.9960.64^AC和PAC模型显著的不为0则该序列存在arch效应。
2) LM 检验：view 下 residual test 下 serial correlation LM testBreuschiTG'odfre/ Serial 'Correlatior LM T&sd!F-stall Stic0.931530Prob.F(2,159Q□：3Q4：0bs*R-^quared1.865551Prob...Chi-Square(2)0.393?Test Equation:Dependentvariable: RESIDMethod: Least SquaresDate: 04/06/11 Time: 15:55Sample:'.2 1602Included observations: 1601Pre^mple missing value lagged residuals set to zero.VariableCoefficientStd. Error t-Btati Stic■■:Prob.LNINDEX(-1)-6.59E-050.001 814 -0.0363350.971C(；3.0004840.013328 0.036322.0.971CRES1D(-1)0.0301 1 80.025082 1^007790.230CRESID(d)-0.0170080.025086 -0.6780030.4975RL$quar已^a.001165Mean dep已「id巳ntx日「7.11E-1EAdjusted R-squared-a.ooo7iiS.D. dependentvar0.01 358SS.E. af regression0.01 3594Akai Ke info criterion-5.75557；Sum ^qyared resid0-2951-'2.3■-Schwarz^riterion-5.74243；Log livelihood461 1.576Hannan-Quinn Writer.-6.750S8^F-stall stir0.621 020Durtiin-Watson stat1.999^7：Prob^F-statistic)0.601417以一个例子为例，主要检验以下变量:F 统计苣： 19.86953 tt), Qb QOOOQGTXR- £1 计贡； 56.57756 概卓隹(F 值儿 (}. 00000()此处的P值为(h拒绝原假般'说明A(6. 1.26)的戟差序列存在ARCH效应"GARCH模型的检验AHJCH-M:3eq.i£ica-tion OptionsV：=d：'i ：=jLc& :iiLii di 5tFibn+ic-Ti sp&c i fi c ：kti 皿 Vari ：u 曲Me：=LTL eqmti on目已卩旨ndu匚lt ty r^^~iS5^rz 也 AfJilA ■terms OK explicitModel: GAHCH/TARCHErt-al (Gjiue £ i srjEquation EstuationTtii- esholdOrder:ARJCH: 1eiLRCH: LRss+ri cti&i NorsMethod 血 hEwtirriyti. on 5^：ttingsAu七oregi~e sei ve Condi i 1 ctl：=l1 Hetero skelasti c 1 kySample： 1 1B02确定 取消Ifll右边的arch-m则为加入均值项，下拉项有两种形式：对数与条件标准差形式均值下面的为' 一‘有几种选项，默认为标准的garch模型,指数garch模型、成分garch模型— 0Ord&r :A^CH: ：f~GAECH: 1表示滞后阶数Do^hoLl 0门槛值的设定，只在指数garch模型的时候才有效V^-i^LCh!. 方差模型的设定，该变量的输入只要求输入“garch定义的方程之外的影响变量（除去残差的方差与滞后n阶的方差）”， 没有则不用输入设定的图为下列：其结果如下:DependenWariable: LNINDE><-Method: ML- ARCH (Marquardt) - Normal distributionDate: 04/06/11 Time: 15<20Sample (adjusted^: 21602Included obsen/atian^: 1601 after adjustmentsCorivergen^e achieved after 1 6 iteration^;Fr■已sam卩I已 variant已：backcast (param已t^r■二 07).GARCH = +C(4yRE3IDC1)A2 + C(5y(?ARGH(-1-')VariableCoefficientStd. Errorz-^Stati-sticProb.LNINDEX^-1)0.9984690.001,254796.46420.0000o.oi figs0.0091981.23913/0.215GVariarice EquationC5 91E-069.85E-075 9984620.0000RESID(-1)A20.1053840.01 01 6E10.363970.0000GARCHMF0.S669440.01071160.940840.0000R/squared0.994789Mean dependentvar7.342821Adjusted R-squareci0.994786S.D. depend&ntv/ar0.108273S.E. of regression0.013595Aka ike infa criterion-5.884161Sum Squared resid0.395542Schwarz criterion-5.867364Log likelihood4715.271Hannan-Quinn enter.-5.877924F-statistic7S31;2'91Durtiin-Watson-stat1.942541Prob(F-statistic>：0.000000解释：上面为均值方程，下面为方差方程。
