发帖
雷达卡
看了很多资料，都提到弱工具变量的检验问题，发现现在比较流行的似乎就是使用ivreg2，使用它报告的Weak identification test (Cragg-Donald Wald F statistic): 1.470
Stock-Yogo weak ID test critical values: 5% maximal IV relative bias 11.04
10% maximal IV relative bias 7.56
20% maximal IV relative bias 5.57
30% maximal IV relative bias 4.73
10% maximal IV size 16.87
15% maximal IV size 9.93
20% maximal IV size 7.54
除此以外，还会报告第一阶段每个内生变量的F值，
F test of excluded instruments:
F( 4, 19) = 44.27
Prob > F = 0.0000
可是如果有两个以上内生变量时，有时候每个的F值都大于10，但是Weak identification test (Cragg-Donald Wald F statistic) 却很小，通不过检验。可是F值说明工具变量和内生变量的相关性已经非常高了，可是为何CD Wald还通不过呢？
不过见过一篇文章是这么做的：
A related question concerns the explanatory power of the instruments. If the instruments are weak the exogenous vari-
ation in openness will be limited and this may distort inference (Stock, Wright, & Yogo, 2002). To address this issue we
regressed DVINFt1 and DOPENit on the instruments used for the differenced equation, and VINFt1 and OPENt on the instru-
ments used for the levels equation, and performed F-tests for the joint significance of the regressors. The test statistics were
27.96 (DVINFt1 equation), 23.16 (DOPENit ), 20.04 (VINFt1) and 83.29 (OPENt ), each of which is significant at the 0.1% level.
Hence, the instruments appear to have considerable explanatory power。
我理解他就是使用了各自的F值来判断弱工具变量问题了。
不知道这样是否可行？
京公网安备 11010802022788号
