[Gretl-users] Average ME

Allin Cottrell cottrell at wfu.edu
Mon Nov 14 07:57:47 EST 2016


On Mon, 14 Nov 2016, J. I. wrote:

> Hey GRETL-USER
>
> May you can help me with a problem.
> I tried to generate AME with the
> script from
>
> http://www.learneconometrics.com/gretl/using_gretl_for_POE4.pdf
>
> p. 379-380, but it wont work.
>
> The Error:
>
> ? ame(&coef, x)
> Liste hat 12 Elemente, aber Länge des Vektors b ist 13
> *** error in function ame, line 1
>
> What means the list is short by one.
>
> List x is all regressors (11) of a ordered probit plus the constant. 
> So it got 12 elements. May it is the DV missig for the Calc? But 
> where to put it?

The dependent variable is not part pf the calculation. However, I'm 
not sure if Lee Adkins' ame() function is set up to handle ordered 
probit. In an ordered model you have one or more "extra" coefficients 
("cut points") which do not correspond to members of the list of 
regressors, and I guess that could account for the mismatch here.

I suspect that one of Oleh Komashko's function packages can handle 
average marginal effects for ordered probit. Oleh?

Allin Cottrell



More information about the Gretl-users mailing list