j***m 发帖数: 74 | 1 if the number of instrument variable is equal to the number of endogeneous
variable (exact-identification), can I still test the exogeneity of the IV?
If so, how? TIA | x********4 发帖数: 405 | 2 yes. ivendog or hausman in stata.
?
【在 j***m 的大作中提到】 : if the number of instrument variable is equal to the number of endogeneous : variable (exact-identification), can I still test the exogeneity of the IV? : If so, how? TIA
| f*******r 发帖数: 257 | 3 This test is to do a Hausman test. It's not to test the endogeneity of the
instruments. It's not to test the endogeneity of the suspected endogenous
variables either. It's a test that given the exogeneity of instruments, do
you really need IV, or you are just fine to do OLS. There is no test to
test endogeneity of instruments. Overidentification test can kind of do
that, but in your case of exact identification, you cannot do that. | d*******n 发帖数: 216 | 4 Even for overidentification test, the test still needs to assume SOME IV is
exogenous and then test whether the rest are exogenous under that assumption
. So fundamentally the exogeneity of IVs can only be argured/reasoned,
instead of being tested.
the
do
【在 f*******r 的大作中提到】 : This test is to do a Hausman test. It's not to test the endogeneity of the : instruments. It's not to test the endogeneity of the suspected endogenous : variables either. It's a test that given the exogeneity of instruments, do : you really need IV, or you are just fine to do OLS. There is no test to : test endogeneity of instruments. Overidentification test can kind of do : that, but in your case of exact identification, you cannot do that.
| h****i 发帖数: 1674 | 5 agree
the exogenousality of the iv can not be totally tested
you must need a story to justify your choice
is
assumption
【在 d*******n 的大作中提到】 : Even for overidentification test, the test still needs to assume SOME IV is : exogenous and then test whether the rest are exogenous under that assumption : . So fundamentally the exogeneity of IVs can only be argured/reasoned, : instead of being tested. : : the : do
|
|