[Gretl-users] A Newbie Needs Help

[Sven Schreiber]

Am 12.03.2008 05:58, Scott David Orr schrieb:
> OK, I've had a look at the gretl documentation on panel models.  I get 
> how to set up panel data, and the basics on panel analysis.  Apparently, 
> if I have time-invariant data (literacy rates would be nearly invariant 
> over the period I'm talking about), I have to use a random-effects 
> model, though that poses a problem if the other IV's are correlated with 
> the panel units (countries, in this case).  I think I can negotiate that 
> problem, though....
> 
> My next question is, how do I handle lags and the need for instruments?  
> The panel documentation mentions this issue, but doesn't seem to go into 
> it, unless it's in another part of the manual.

Well actually I haven't worked yet with the panels&gretl combination, so 
  I can't help you with that. Dynamic panels are not trivial (but 
fashionable), maybe you want to use something like the Arrellano/Bond 
estimator. Of course you could perform that by cleverly using gretl's 
GMM framework, but I think it's not natively implemented yet. (Somebody 
correct me please.)

good luck,
sven