金融数据处理(波动溢出).doc

金融数据处理(波动溢出)由于Eviews在对波动溢出效应估计方面，只能得到对角矩阵。因此在用BEKK模型时，将使用S-Plus配合来估计波动溢出效应。实验步骤：第一步 检查平稳性第二步 协整检验第三步 Granger因果检验第四步 VAR第五步 VEC第六步 BEKK第一步 检查平稳性Null Hypothesis: RHIS has a unit rootExogenous: Constant, Linear TrendLag Length: 0 (Automatic based on SIC, MAXLAG=21)t-StatisticProb.*Augmented Dickey-Fuller test statistic-35.207310.0000Test critical values:1% level-3.9662335% level-3.41381510% level-3.128983*MacKinnon (1996) one-sided p-values.Augmented Dickey-Fuller Test EquationDependent Variable: D(RHIS)Method: Least SquaresDate: 11/20/11 Time: 16:28Sample (adjusted): 1/06/2006 11/08/2010Included observations: 1124 after adjustmentsVariableCoefficientStd. Errort-StatisticProb.RHIS(-1)-1.0502140.029829-35.207310.0000C0.0004970.0012360.4022690.6876TREND(1/04/2006)-6.72E-081.90E-06-0.0353410.9718R-squared0.525112Mean dependent var-1.09E-06Adjusted R-squared0.524264S.D. dependent var0.029972S.E. of regression0.020673Akaike info criterion-4.917329Sum squared resid0.479076Schwarz criterion-4.903918Log likelihood2766.539Hannan-Quinn criter.-4.912261F-statistic619.7774Durbin-Watson stat1.999385Prob(F-statistic)0.000000可知，rhis系列没有单位根，即是平稳的。Null Hypothesis: RNQI has a unit rootExogenous: Constant, Linear TrendLag Length: 0 (Automatic based on SIC, MAXLAG=21)t-StatisticProb.*Augmented Dickey-Fuller test statistic-36.799060.0000Test critical values:1% level-3.9662335% level-3.41381510% level-3.128983*MacKinnon (1996) one-sided p-values.Augmented Dickey-Fuller Test EquationDependent Variable: D(RNQI)Method: Least SquaresDate: 11/20/11 Time: 16:27Sample (adjusted): 1/06/2006 11/08/2010Included observations: 1124 after adjustmentsVariableCoefficientStd. Errort-StatisticProb.RNQI(-1)-1.0941330.029733-36.799060.0000C-0.0004430.001071-0.4133050.6795TREND(1/04/2006)1.00E-061.65E-060.6096650.5422R-squared0.547102Mean dependent var-5.67E-06Adjusted R-squared0.546294S.D. dependent var0.026591S.E. of regression0.017911Akaike info criterion-5.204124Sum squared resid0.359626Schwarz criterion-5.190713Log likelihood2927.718Hannan-Quinn criter.-5.199056F-statistic677.0856Durbin-Watson stat2.010733Prob(F-statistic)0.000000可见，rnqi是平稳的Null Hypothesis: RSSE has a unit rootExogenous: Constant, Linear TrendLag Length: 0 (Automatic based on SIC, MAXLAG=21)t-StatisticProb.*Augmented Dickey-Fuller test statistic-33.951420.0000Test critical values:1% level-3.9662335% level-3.41381510% level-3.128983*MacKinnon (1996) one-sided p-values.Augmented Dickey-Fuller Test EquationDependent Variable: D(RSSE)Method: Least SquaresDate: 11/20/11 Time: 16:29Sample (adjusted): 1/06/2006 11/08/2010Included observations: 1124 after adjustmentsVariableCoefficientStd. Errort-StatisticProb.RSSE(-1)-1.0139250.029864-33.951420.0000C0.0026660.0012762.0890620.0369TREND(1/04/2006)-3.18E-061.96E-06-1.6203640.1054R-squared0.506971Mean dependent var-3.88E-06Adjusted R-squared0.506092S.D. dependent var0.030314S.E. of regression0.021304Akaike info criterion-4.857180Sum squared resid0.508777Schwarz criterion-4.843769Log likelihood2732.735Hannan-Quinn criter.-4.852112F-statistic576.3509Durbin-Watson stat2.000314Prob(F-statistic)0.000000可见，rsse序列是平稳的。第二步 协整检验Date: 11/20/11 Time: 16:30Sample (adjusted): 1/12/2006 11/08/2010Included observations: 1120 after adjustmentsTrend assumption: Linear deterministic trendSeries: RHIS RNQI RSSELags interval (in first differences): 1 to 4Unrestricted Cointegration Rank Test (Trace)HypothesizedTrace0.05No. of CE(s)EigenvalueStatisticCritical ValueProb.*None *0.222478694.669929.797070.0001At most 1 *0.185719412.829015.494710.0001At most 2 *0.150534182.72473.8414660.0000Trace test indicates 3 cointegrating eqn(s) at the 0.05 level* denotes rejection of the hypothesis at the 0.05 level*MacKinnon-Haug-Michelis (1999) p-valuesUnrestricted Cointegration Rank Test (Maximum Eigenvalue)HypothesizedMax-Eigen0.05No. of CE(s)EigenvalueStatisticCritical ValueProb.*None *0.222478281.840921.131620.0001At most 1 *0.185719230.104314.264600.0001At most 2 *0.150534182.72473.8414660.0000Max-eigenvalue test indicates 3 cointegrating eqn(s) at the 0.05 level* denotes rejection of the hypothesis at the 0.05 level*MacKinnon-Haug-Michelis (1999) p-valuesUnrestricted Cointegrating Coefficients (normalized by b*S11*b=I):RHISRNQIRSSE147.6973-189.3795-40.1063793.0083739.50010-2.556401-15.00984-48.00880108.6994Unrestricted Adjustment Coefficients (alpha):D(RHIS)-0.004785-0.006733-0.003038D(RNQI)0.005305-0.0069690.000477D(RSSE)0.001030-0.001374-0.0086351 Cointegrating Equation(s):Log likelihood8569.234Normalized cointegrating coefficients (standard error in parentheses)RHISRNQIRSSE1.000000-1.282214-0.271544(0.05596)(0.04073)Adjustment coefficients (standard error in parentheses)D(RHIS)-0.706744(0.08719)D(RNQI)0.783598(0.08442)D(RSSE)0.152080(0.10022)2 Cointegrating Equation(s):Log likelihood8684.286Normalized cointegrating coefficients (standard error in parentheses)RHISRNQIRSSE1.0000000.000000-0.088210(0.05379)0.0000001.0000000.142983(0.04732)Adjustment coefficients (standard error in parentheses)D(RHIS)-1.3330030.640229(0.09679)(0.10728)D(RNQI)0.135458-1.280001(0.09282)(0.10288)D(RSSE)0.024311-0.249262(0.11822)(0.13103)由于prob很小，说明rhis , rnqi , rsse是协整的，具有长期的均衡关系。第三步 Granger因果检验Pairwise Granger Causality TestsDate: 11/20/11 Time: 16:33Sample: 1/04/2006 11/08/2010Lags: 5Null Hypothesis:ObsF-StatisticProb.RNQI does not Granger Cause RHIS112062.74391.E-57RHIS does not Granger Cause RNQI0.535950.7492RSSE does not Granger Cause RHIS11202.999940.0107RHIS does not Granger Cause RSSE1.432820.2097RSSE does not Granger Cause RNQI11202.343450.0396RNQI does not Granger Cause RSSE8.424518.E-08可知，RNQI是RHIS的Granger原因，RHIS不是RNQI的Granger原因； RSSE是RHIS的Granger原因，RHIS不是RSSE的Granger原因； RSSE是RNQI的Granger原因，RNQI是RSSE的Granger原因。第四步 VARVector Autoregression EstimatesDate: 11/20/11 Time: 16:38Sample (adjusted): 1/09/2006 11/08/2010Included observations: 1123 after adjustmentsStandard errors in ( ) & t-statistics in RHISRNQIRSSERHIS(-1)-0.2114520.002593-0.028258(0.03487)(0.03401)(0.03993)-6.06457 0.07622-0.70777RHIS(-2)-0.0147010.048541-0.011387(0.03124)(0.03047)(0.03577)-0.47064 1.59292-0.31836RNQI(-1)0.558608-0.1011990.228046(0.03247)(0.03167)(0.03718) 17.2061-3.19530 6.13428RNQI(-2)0.179709-0.0752440.050368(0.03608)(0.03520)(0.04131) 4.98108-2.13787 1.21919RSSE(-1)-0.0600390.021830-0.020169(0.02919)(0.02847)(0.03342)-2.05708 0.76670-0.60348RSSE(-2)-0.009692-0.020194-0.011206(0.02920)(0.02848)(0.03344)-0.33192-0.70892-0.33515C0.0005089.74E-050.000868(0.00055)(0.00053)(0.00063) 0.92586 0.18206 1.38193R-squared0.2181830.0149940.034557Adj. R-squared0.2139800.0096980.029366Sum sq. resids0.3754820.3573310.492331S.E. equation0.0183430.0178940.021004F-statistic51.907452.8313816.657596Log likelihood2900.3872928.2082748.254Akaike AIC-5.152960-5.202508-4.882019Schwarz SC-5.121645-5.171194-4.850704Mean dependent0.0004330.0001000.000855S.D. dependent0.0206890.0179810.021319Determinant resid covariance (dof adj.)3.36E-11Determinant resid covariance3.30E-11Log likelihood8771.514Akaike information criterion-15.58417Schwarz criterion-15.49023第五步 VECVector Error Correction EstimatesDate: 11/20/11 Time: 16:47Sample (adjusted): 1/10/2006 11/08/2010Included observations: 1122 after adjustmentsStandard errors in ( ) & t-statistics in Cointegrating Eq:CointEq1RHIS(-1)1.000000RNQI(-1)-1.112045(0.03985)-27.9026RSSE(-1)-0.144771(0.03056)-4.73772C-0.000197Error Correction:D(RHIS)D(RNQI)D(RSSE)CointEq1-1.1210890.465546-0.166621(0.06299)(0.06533)(0.07726)-17.7978 7.12577-2.15672D(RHIS(-1)-0.059345-0.3659190.094168(0.04799)(0.04977)(0.05886)-1.23664-7.35172 1.59993D(RHIS(-2)-0.045722-0.1806740.060795(0.03023)(0.03136)(0.03708)-1.51224-5.76146 1.63944D(RNQI(-1)-0.521618-0.3327320.059792(0.05966)(0.06188)(0.07317)-8.74330-5.37726 0.81715D(RNQI(-2)-0.193842-0.1900970.065966(0.03800)(0.03942)(0.04661)-5.10055-4.82267 1.41522D(RSSE(-1)-0.1403020.071703-0.708557(0.02567)(0.02662)(0.03148)-5.46593 2.69327-22.5067D(RSSE(-2)-0.0707610.026966-0.388488(0.02541)(0.02636)(0.03117)-2.78456 1.02310-12.4646C-1.14E-05-1.19E-066.59E-06(0.00058)(0.00061)(0.00072)-0.01956-0.00197 0.00922R-squared0.5783360.4236570.379938Adj. R-squared0.5756860.4200360.376041Sum sq. resids0.4253530.4575760.639844S.E. equation0.0195400.0202670.023966F-statistic218.2734116.982597.51339Log likelihood2827.3432786.3772598.283Akaike AIC-5.025566-4.952544-4.617261Schwarz SC-4.989753-4.916730-4.581447Mean dependent-8.58E-06-3.19E-063.86E-06S.D. dependent0.0299980.0266130.030340Determinant resid covariance (dof adj.)5.67E-11Determinant resid covariance5.55E-11Log likelihood8471.907Akaike information criterion-15.05331Schwarz criterion-14.93244第六步 VECH rhis rnqi rsse rtn.matrix bekk=mgarch(rtn.matrix1,bekk(1,1) summary(bekk)Call: mgarch(formula.mean = rtn.matrix 1, formula.var = bekk(1, 1)Mean Equation: structure(.Data = rtn.matrix 1, class = formula)Conditional Variance Equation: structure(.Data = bekk(1, 1), class = formula)Conditional Distribution: gaussian-Estimated Coefficients:- Value Std.Error t value Pr(|t|) C(1) 0.00099467 0.0004234 2.34899 1.900e-002 C(2) 0.00084144 0.0004058 2.07352 3.835e-002 C(3) 0.00132681 0.0005421 2.44734 1.454e-002 A(1, 1) 0.00257109 0.0003719 6.91263 7.987e-012 A(2, 1) 0.00002574 0.0004335 0.05937 9.527e-001 A(3, 1) -0.00024515 0.0007504 -0.32669 7.440e-001 A(2, 2) 0.00090826 0.0004504 2.01644 4.399e-002 A(3, 2) -0.00049733 0.0013717 -0.36257 7.170e-001 A(3, 3) 0.00228036 0.0005070 4.49806 7.575e-006 ARCH(1; 1, 1) 0.33970572 0.0243489 13.95160 0.000e+000 ARCH(1; 2, 1) -0.04826711 0.0248394 -1.94316 5.225e-002 ARCH(1; 3, 1) 0.06403003 0.0260252 2.46031 1.403e-002 ARCH(1; 1, 2) 0.10497113 0.0239241 4.38767 1.254e-005 ARCH(1; 2, 2) 0.25802307 0.0148418 17.38493 0.000e+000 ARCH(1; 3, 2) -0.03952891 0.0234679 -1.68439 9.239e-002 ARCH(1; 1, 3) -0.08230638 0.0186652 -4.40960 1.136e-005 ARCH(1; 2, 3) 0.00800872 0.0180357 0.44405 6.571e-001 ARCH(1; 3, 3) 0.20037091 0.0208436 9.61308 0.000e+000GARCH(1; 1, 1) 0.90860992 0.0101064 89.90434 0.000e+000GARCH(1; 2, 1) -0.03328475 0.0096157 -3.46151 5.575e-004GARCH(1; 3, 1) -0.01712419 0.0107774 -1.58890 1.124e-001GARCH(1; 1, 2) 0.03790046 0.0074664 5.07616 4.509e-007GARCH(1; 2, 2) 0.97637454 0.0047567 205.26361 0.000e+000GARCH(1; 3, 2) 0.03399328 0.0046182 7.36074 3.539e-013GARCH(1; 1, 3) 0.03192512 0.0077946 4.09582 4.511e-005GARCH(1; 2, 3) 0.00259084 0.0059258 0.43722 6.620e-001GARCH(1; 3, 3) 0.97294599 0.0070750 137.51922 0.000e+000-AIC(27) = -18257.38BIC(27) = -18121.69Normality Test:- Jarque-Bera P-value Shapiro-Wilk P-value Col1 170.07 0 0.9865 4.156e-001Col1 2211.46 0 0.9636 0.000e+000Col1 94.83 0 0.9746 8.362e-011Ljung-Box test for standardized residuals:- Statistic P-value Chi2-d.f. Col1 6.869 8.661e-001 12Col1 47.309 4.121e-006 12C