我在做一个问题时编了如下程序:
proc autoreg data=szzz;
model cl = date / nlag=1 garch=(q=1,p=1) maxit=50;
hetero d;
output out=szzzv cev=vhat;
run;
结果显示:
GARCH Estimates
SSE 706317.636 Observations 1210
MSE 583.73358 Uncond Var 6.92482E-7
Log Likelihood -5309.3859 Total R-Square 0.9960
SBC 10661.3621 AIC 10630.7719
Normality Test 328.9283 Pr > ChiSq <.0001
Standard Approx Variable
Variable DF Estimate Error t Value Pr > |t| Label
Intercept 1 -6070 4247 -1.43 0.1529
date 1 0.4975 0.2588 1.92 0.0546 F1
AR1 1 -0.9998 0.000775 -1290.8 <.0001
ARCH0 1 1.0537E-8 1.152E-14 914275 <.0001
ARCH1 1 0.0928 0.0118 7.85 <.0001
GARCH1 1 0.8920 0.0127 70.22 <.0001
HET1 1 42.6263 10.8113 3.94 <.0001
但加了mean选项后,即:
proc autoreg data=szzz;
model cl = date / nlag=1 garch=(q=1,p=1,mean) maxit=50;
hetero d;
output out=szzzv cev=vhat;
run;
结果:
GARCH Estimates
SSE 689615.951 Observations 1210
MSE 569.93054 Uncond Var .
Log Likelihood -5307.1043 Total R-Square 0.9961
SBC 10670.9956 AIC 10630.2086
Normality Test 302.2562 Pr > ChiSq <.0001
Standard Approx Variable
Variable DF Estimate Error t Value Pr > |t| Label
Intercept 1 -6070 4737 -1.28 0.2000
date 1 0.4958 0.2964 1.67 0.0944 F1
AR1 1 -0.9996 0.000944 -1058.7 <.0001
ARCH0 1 3.8993 5.4715 0.71 0.4761
ARCH1 1 0.0915 0.0128 7.15 <.0001
GARCH1 1 0.8966 0.0140 63.99 <.0001
DELTA 1 1.1014 0.5072 2.17 0.0299
HET1 1 17.8203 24.5416 0.73 0.4678
为什么会影响到GARCH方程的参数估计值呢?
请教了!!!