四，均值方程的确定及残差系列自相关检验
ls c2 c c2(-1)
|
Variable |
Coefficient |
Std. Error |
t-Statistic |
Prob. |
|
C |
0.005056 |
1.583308 |
0.003193 |
0.9975 |
|
C2(-1) |
0.138123 |
0.235657 |
0.586117 |
0.5696 |
ls c2 c c2(-2)
|
Variable |
Coefficient |
Std. Error |
t-Statistic |
Prob. |
|
C |
0.101599 |
1.710089 |
0.059412 |
0.9538 |
|
C2(-2) |
-0.113372 |
0.247135 |
-0.458746 |
0.6562 |
ls c2 c c2(-3)
|
Variable |
Coefficient |
Std. Error |
t-Statistic |
Prob. |
|
C |
-0.370311 |
1.810253 |
-0.204563 |
0.8425 |
|
C2(-3) |
-0.100433 |
0.251182 |
-0.399841 |
0.6986 |
ls c2 c c2(-4)
|
Variable |
Coefficient |
Std. Error |
t-Statistic |
Prob. |
|
C |
0.172654 |
1.527052 |
0.113064 |
0.9128 |
|
C2(-4) |
-0.504583 |
0.202425 |
-2.492697 |
0.0374 |
ls c2 c c2(-5)
|
Variable |
Coefficient |
Std. Error |
t-Statistic |
Prob. |
|
C |
1.237494 |
0.758823 |
1.630806 |
0.1470 |
|
C2(-5) |
-0.100520 |
0.095914 |
-1.048030 |
0.3295 |
因此c2与它的四阶滞后存在显著的自相关，因此c2的均值方程存在如下形式：
c2=c+ a* c2(-4) + e
对e进行自相关检验:
ls c2 c c2(-4)
genr e=resid()
|
e |
ac |
-0.006 |
-0.478 |
0.014 |
0.12 |
-0.09 |
0.185 |
0.091 |
-0.249 |
|
p值 |
0.984 |
0.181 |
0.331 |
0.446 |
0.563 |
0.553 |
0.628 |
0.344 |
|
E^2 |
ac |
-0.259 |
0.312 |
-0.18 |
0.217 |
-0.21 |
-0.09 |
-0.24 |
0.106 |
|
p值 |
0.345 |
0.309 |
0.406 |
0.427 |
0.431 |
0.527 |
0.39 |
0.428 |
line e*e
所以e^2有明显的时间可变性和集族性,适合用GARCH模型.
对e进行ARCH---LM Test有:
一阶:
|
ARCH Test: | |||
|
F-statistic |
2.055004 |
Probability |
0.194823 |
|
Obs*R-squared |
2.042521 |
Probability |
0.152956 |
两阶:
|
ARCH Test: | |||
|
F-statistic |
2.203858 |
Probability |
0.205928 |
|
Obs*R-squared |
3.748171 |
Probability |
0.153495 |
三阶:
|
ARCH Test: | |||
|
F-statistic |
0.244597 |
Probability |
0.861148 |
|
Obs*R-squared |
1.375688 |
Probability |
0.711243 |
ARCH效应还算明显.
评论