[Gretl-users] maximum like. conditions

Allin Cottrell cottrell at wfu.edu
Sun Nov 29 12:06:38 EST 2009


On Sun, 29 Nov 2009, Berend Hasselman wrote:

> On 29-11-2009, at 15:24, Davor Horvatic wrote:
>
> > OK, I figured out the correct code, so please confirm that I interpret
> > correctly.
> > Restriction criteria has to be written as
> >
> > scalar check = (alpha>0) && (beta>0)
> >
>
> That is what I tried and got the error message
>
> Data types not conformable for operation
>
> Error executing script: halting
>
> If it does work for you, something weird is going on.
>
> I am running  Gretl 1.8.6 on Mac OS X 10.6.2.
>
> We'll have to wait for the Gretl guru to have a look at this.

;-) This is a subtle bug (if it's a bug; I'm not sure) with an
easy workaround.  With the initialization

scalar alpha = 0.4
scalar beta = 0

The "check" condition -- alpha>0 && beta>0 -- is violated on the
first iteration.  Therefore the formula for "ll" comes down to

ll = NA

This generates a scalar value, which is not allowed in the mle
context.  The fix is to initialize such that the check is
satisfied on the first round, e.g.

scalar beta = 0.001

Even if one didn't get a scalar for ll, I don't think that mle
would work with the constraint not satisfied at the start.  The
first thing we do is use the ll formula to check for missing
values, and that can't work if the entire series is NAs.

Allin.


More information about the Gretl-users mailing list