y****n
发帖数: 2 | 1 看到一篇文章里写到：“identifying a high-quality instrument is extremely
challenging both because it is often difficult to find an instrument that is
conceptually plausible and because there is no analytical method for
testing the quality of an instrument.....
又写道：
A variable must satisfy two criteria if it is to serve as a valid instrument
for Varible X by Variable Y. First, it must strongly predict Variable X.
Second, net of its effect on Variable X, it should be unrelated to Varible Y.
The first criteri | x********4
发帖数: 405 | 2 that is generally true. as far as i know, you need to have more IVs than
endogenous variables (overidentification) and use some sargan or basmann's
tests to examine the joint null hypothesis that the excluded IVs are valid
ones, but we really do not have direct test to examine the second criteria...
is
instrument
Y.
examining
【在 y****n 的大作中提到】
: 看到一篇文章里写到：“identifying a high-quality instrument is extremely
: challenging both because it is often difficult to find an instrument that is
: conceptually plausible and because there is no analytical method for
: testing the quality of an instrument.....
: 又写道：
: A variable must satisfy two criteria if it is to serve as a valid instrument
: for Varible X by Variable Y. First, it must strongly predict Variable X.
: Second, net of its effect on Variable X, it should be unrelated to Varible Y.
: The first criteri
|
|
