[Gretl-users] Re: How to exploit the panel data in a SEM
Am 12.09.2019 um 14:20 schrieb luis.salazar.ramirez:
Good day for all.
I have a SEM where the structure is panel data.
I load the data in gretl, I find options to manage panel data but in the
SEM I can't find some option relative to panel data. Is Gretl
exploiting the panel data estructure already ? or I have to center the
data before to load data on, to manage fixed effect, for example?.
Hi, indeed I am not aware of specific methods in gretl to estimate
systems of simultaneous equations (SEM) on panel data per se. As the doc
for the 'system' command says, you have the generic estimation
possibilities of ols, tsls, sur, 3sls, fiml.
As you say, to do fixed effects you could simply apply the within
transformation (y - pmean(y)) to all variables. However, I'd say that
this would be problematic in a system context because of the strict
exogeneity assumption of FE. If x2 is endogenous in equation 2, you
cannot defend its exogeneity as a regressor in equ. 1 in the same
system. Maybe you should also take a look at the orthdev() function.
I guess one natural way would be to use IV/GMM-based approaches in the
Arellano-Bond tradition, as in gretl's 'dpanel' command. As I see it, in
gretl you would simply specify separate equations, estimate them
separately, and them consider them together as a system. In the
literature there are probably SEM-type extensions of Arellano-Bond, but
these quite specialized things AFAIK are not in gretl.
Another option for large-T (and small-N) panels might be to simply use
pure time series methods.
cheers
sven