我在高级视频 B6_TimeS 中讲解了这一问题。
*== 脉冲响应 == (第 810 行)
我提到进行脉冲响应分析分为三个步骤:
* -- Basic steps
* step1: estimate VAR model
use lutkepohl.dta, clear
var dlinvestment dlincome dlconsumption,lag(1/2) dfk small
* step2: create IRF file
irf create order1, step(10) set(myirf1) replace
* step3: graphing
irf graph oirf, impulse(dlincome) response(dlconsumption) irf(order1)
这里的第三步可以绘制出脉冲响应图,每个单独的图形中都会有一条灰色的带子,这就是相应脉冲相应的置信区间,用于检验脉冲响应的显著程度。如果这个带子穿过了横轴 (即 y = 0),则表明置信区间包含零,意味着脉冲响应不显著。
你也可以采用表格的方式呈现脉冲相应的结果:
*-- show results in table format (第 850 行)
irf table oirf, irf(order1 order2) impulse(dlincome) response(dlconsumption)
结果如下:
+--------------------------------------------------------------------------------+
| | (1) (1) (1) | (2) (2) (2) |
| step | oirf Lower Upper | oirf Lower Upper |
|--------+-----------------------------------+-----------------------------------|
|0 | .005338 .003545 .00713 | .00569 .003802 .007578 |
|1 | .001704 -.000385 .003792 | .001644 -.000452 .003739 |
|2 | .003071 .000963 .005179 | .003445 .00124 .00565 |
|3 | -.00023 -.001636 .001176 | -.000198 -.001635 .001239 |
|4 | .000845 -.000402 .002092 | .000881 -.000432 .002193 |
|5 | .000481 -.000227 .001189 | .000512 -.000229 .001253 |
|6 | .000045 -.000539 .00063 | .000072 -.000548 .000692 |
|7 | .000157 -.000187 .000502 | .000165 -.0002 .00053 |
|8 | .000095 -.000148 .000338 | .0001 -.000159 .000359 |
|9 | .000019 -.000142 .00018 | .000024 -.000147 .000196 |
|10 | .000036 -.000065 .000136 | .000038 -.00007 .000146 |
+--------------------------------------------------------------------------------+
95% lower and upper bounds reported
(1) irfname = order1, impulse = dlincome, and response = dlconsumption
(2) irfname = order2, impulse = dlincome, and response = dlconsumption
Lower 和 Upper 分别是 95% 置信区间的下限和上限,如果这两个值均为正值或负值,则表明脉冲响应显著为正或显著为负。
例如 [.003545, .00713] 便表明第 0 期的脉冲相应显著为正。而 [-.000385, .003792] 则意味着第1期的脉冲响应不显著。