Re: [Gretl-users] limitations of maximum likelihood estimations
On Sun, February 19, 2006 18:33, Allin Cottrell wrote:
On Sun, 19 Feb 2006, Arnaud Bervas wrote:
> I have also tried with artificial data, but still doesn't work. I
> was about to reinstall Gretl to try again, but you will maybe find
> out what is wrong using the data attached. Thank you for this
I've run the script with your data, no problem. But I think perhaps
you were defining the MLE model in the gretl's gui MLE window -- is
that right? I have discovered a bug -- now fixed in CVS -- with
long expressions that include line breaks in that window.
This fix will be in the next release. In the mantime, if you
haven't already done so, try running your example in script mode.
Or, perhaps, tidy up the syntax a little:
mle loglik= log (p3*((1-p1)*x1+p1*(x2)) +(1-p3)*x3)
scalar p1 = 1/(1+exp(-c1))
scalar p2 = 1/(1+exp(-c2))
scalar p3 = 1/(1+exp(-c3))
series x1 = (0.5*(1+p2))^BUY *(0.5*(1-p2))^SELL
series x2 = (0.5*(1-p2))^BUY *(0.5*(1+p2))^SELL
series x3 = 0.5^(BUY+SELL)
params c1 c2 c3
end mle
Not only you avoid long lines, you also have a much more readable script;
further optimisations are possible, but I'm sure you get the idea.
--
Riccardo "Jack" Lucchetti
Dipartimento di Economia
Facoltà di Economia "G. Fuà"
Ancona