我要对期货和现货套期保值,数据取对数后是进行单位根检验
Null Hypothesis: SERIES01 has a unit root
Exogenous: Constant
Lag Length: 3 (Automatic - based on SIC, maxlag=24)
t-Statistic Prob.*
Augmented Dickey-Fuller test statistic -1.287300 0.6375
Test critical values: 1% level -3.434013
5% level -2.863045
10% level -2.567619
数据不平稳,然后就取了一阶差分。
之后的各种模型OLS,GARCH求套保比率时结果就放飞自我了,这是一阶差分后GARCH(1,1)的结果
Dependent Variable: SERIES01
Method: ML ARCH - Normal distribution (BFGS / Marquardt steps)
Date: 01/24/18 Time: 21:08
Sample: 1 1694
Included observations: 1694
Convergence achieved after 38 iterations
Coefficient covariance computed using outer product of gradients
Presample variance: backcast (parameter = 0.7)
GARCH = C(3) + C(4)*RESID(-1)^2 + C(5)*GARCH(-1)
Variable Coefficient Std. Error z-Statistic Prob.
SERIES02 -0.057837 0.001141 -50.70190 0.0000
C 0.000482 1.42E-05 33.80450 0.0000
Variance Equation
C 6.52E-07 4.70E-08 13.87213 0.0000
RESID(-1)^2 4.043876 0.177559 22.77481 0.0000
GARCH(-1) 0.079430 0.003663 21.68711 0.0000
R-squared -0.046723 Mean dependent var 3.54E-05
Adjusted R-squared -0.047341 S.D. dependent var 0.005422
S.E. of regression 0.005549 Akaike info criterion -8.917948
Sum squared resid 0.052091 Schwarz criterion -8.901906
Log likelihood 7558.502 Hannan-Quinn criter. -8.912008
Durbin-Watson stat 1.250858
我觉得这个肯定不对,然后用对数价格lnP和lnF计算了一下,得到的套期保值比率是0.8
Dependent Variable: SERIES03
Method: ML ARCH - Normal distribution (BFGS / Marquardt steps)
Failure to improve likelihood (singular hessian) after 33 iterations
Coefficient covariance computed using outer product of gradients
Presample variance: backcast (parameter = 0.7)
GARCH = C(3) + C(4)*RESID(-1)^2 + C(5)*GARCH(-1)
Variable Coefficient Std. Error z-Statistic Prob.
SERIES04 0.831364 0.001075 773.4571 0.0000
C 1.633407 0.010499 155.5725 0.0000
Variance Equation
C 1.80E-05 2.42E-06 7.430144 0.0000
RESID(-1)^2 0.840052 0.065298 12.86492 0.0000
GARCH(-1) 0.230778 0.031903 7.233822 0.0000
R-squared 0.912664 Mean dependent var 9.767711
Adjusted R-squared 0.912612 S.D. dependent var 0.225065
S.E. of regression 0.066533 Akaike info criterion -3.967404
Sum squared resid 7.489775 Schwarz criterion -3.951362
Log likelihood 3365.391 Hannan-Quinn criter. -3.961464
Durbin-Watson stat 0.043144
有大神能看出问题在哪里嘛?
非常不明白价格的相关性能达到0.9以上的数据取对数收益率后相关性就0.1左右,可不可以用对数价格lnP做计算
附件列表