[Gretl-users] are TSLS-related tests accessible after estimation?

[Sven Schreiber]

Riccardo (Jack) Lucchetti schrieb:
> On Thu, 31 Jan 2008, Sven Schreiber wrote:
> 
>> So either make things like $hausman and $sargan matrices with two or 
>> three elements each: test statistic, possibly d.o.f., p-value.
> 
> In my current source tree, you now have that if you run
> 
> <script>
> open denmark
> tsls 1 0 2 ; 0 3 4
> a = $hausman
> print a
> </script>
> 
> you get
> 
> <output>
> Model 1: TSLS estimates using the 55 observations 1974:1-1987:3
> Dependent variable: LRM
> Instruments: IBO IDE
> 
>       VARIABLE       COEFFICIENT        STDERROR      T STAT   P-VALUE
> 
>   const                -7.01705          2.60634      -2.692   0.00710 ***
>   LRY                   3.15360          0.437855      7.202  <0.00001 ***
> 
> [...snip...]
> 
> Hausman test -
>   Null hypothesis: OLS estimates are consistent
>   Asymptotic test statistic: Chi-square(1) = 101.11
>   with p-value = 8.69944e-24
> 
> [...snip...]
> 
> ? a = $hausman
> Generated matrix a
> ? print a
> a (1 x 3)
> 
>       101.11       1.0000   8.6994e-24
> </output>
> 
> Would this be ok? If it is, I'll add $sargan too and submit a patch to 
> Allin for review.
> 

For me this would be fine. The rest is for you and Allin to decide...

thanks,
sven