[Gretl-users] Johansen

[Riccardo (Jack) Lucchetti]

On Thu, December 21, 2006 15:00, Sven Schreiber wrote:
> javier garcia enriquez schrieb:
>
>> r>r0), while others defend the following:
>>
>> Trace                            H0: r  <=  r0        H1: r>r0
>> Lambda max                 H0: r =r0           H1: r>r0
>>
>
> As John already answered 2 months ago, the null hypotheses are the same
> in principle. And as I answered 2 months ago, whether you posit <=0 or
> =0 depends on your interpretation or previous test results. (Actually,
> one could also argue that the alternative hypotheses are the same, but
> that would probably cause still more confusion...)
>
>>
>> Is Gretl use either of them??
>>
>
> Gretl's output is very clear about this (it reports both), so where do
> your doubts come from?
>

In my experience, a nice way to put this is to think of the trace and
\lambda_{max} tests as test for zero eigenvalues: the difference is that the
\lambda_{max} test checks if THAT PARTICULAR eigenvalue is 0, while the trace
test checks if all eigenvalues from a certain point on are 0 *jointly*. It's a
bit like a t-test vs an F-test.


Riccardo (Jack) Lucchetti
Dipartimento di Economia
Facoltà di Economia "G. Fuà"
Ancona