Riccardo (Jack) Lucchetti schrieb:
Could you be more specific on this? I mean, suppose you want to estimate
something like y_t = f(x_t, \theta) + \epsilon_t via NLIV. What you can
do with GMM is write the orthogonality condition as y_t - f(x_t, \theta)
and go from there.
Well I can't be more specific, your formulation already captures all of
it AFAICT. Only that it will be a problem of endogeneity bias in a
single equation, not measurement error, but that shouldn't really affect
the usage.
A little example script follows, where I set up a simple example of NLIV
estimation in a measurement error textbook situation:
See the point is I can't even tell yet whether that is what I need
because I don't know the details of gretl's gmm syntax. For example,
without too much thinking I would expect some derivatives to be needed
(X(\tilde(\beta)) in Davidson-MacKinnon-speak), but maybe that's already
implicitly there?
Apart from that your suggestion is probably very useful, so thanks alot.
-Sven