On Mon, 21 Dec 2015, Raul Gimeno wrote:

> Dear all
>
>
>
> Is there a possibility to perform this test manually?
>
> Step 1: Rank the observations according to the values to Xi, beginning with
> the lowest value
>
> Step 2: Omit c central observations
>
> Step 3: Fit separate OLS regressions to the first (n-c)/2 observations and
> the last (n-c)/2 observations
>

For the edification of the youth:

<hansl>
open mroz87.gdt --quiet

list X = WA WE KL6
smpl WW X --no-missing --permanent

scalar cutpoint_lo = quantile(WA, 1/3)
scalar cutpoint_hi = quantile(WA, 2/3)

ols WW 0 X
smpl WA<cutpoint_lo --restrict
scalar den = $ess / $df
scalar df1 = $df
ols WW 0 X --quiet
smpl WA>=cutpoint_hi --restrict --replace
ols WW 0 X --quiet
scalar num = $ess / $df
scalar df2 = $df
smpl full
scalar GT = num/den
pvalue F df2 df1 GT
</hansl>

>
> Would it be possible to complement gretl menu with this test?
>

I'd rather not. Nobody uses this test anymore in serious empirical 
research, and it's rapidly disappearing from textbooks as well. Like Oleh 
said, you're welcome to submit a function package if you're into 
antiquities.

-------------------------------------------------------
   Riccardo (Jack) Lucchetti
   Dipartimento di Scienze Economiche e Sociali (DiSES)

   Università Politecnica delle Marche
   (formerly known as Università di Ancona)

   r.lucchetti@univpm.it
   http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------