Am 18.11.2013 14:25, schrieb Matthieu Stigler:
I tried to use the IV-LIML estimator on a super simple model that does
not include any constant (both in regressors and instruments). However,
when trying to do so, I get an error message, like "data error"
(actually: "erreur de donnée" in the french version). It seems however
possible to run the same model under 2SLS.
Is this on purpose? If not, maybe the code could be adjusted? I has
probably something to do with the fact that the first M projection
matrix becomes an identity under such circumstances?
Slightly OT: Can there be a good reason not to include the constant as
an instrument? Perhaps for pedagogical purposes, but in reality?