Re: [Gretl-devel] Chow and QLR tests
On Fri, 28 Dec 2012, Sven Schreiber wrote:
On 12/28/2012 05:37 PM, Allin Cottrell wrote:
> A user has pointed out to me that gretl's definition of the break
> point in the Chow and QLR tests is not the majority definition. That
> is, when we do a Chow test with "break at observation tau", the
> dummies we create kick in at tau. This is certainly defensible, but
> the more standard procedure is to make the dummies kick in at tau+1:
> tau is then the last observation of regime 1 rather than the first
> observation of regime 2.
>
> I'm wondering if we should change this (and document it), or maybe
> just document clearly what we do at present.
>
> Any thoughts?
>
Hi,
hope you had happy holidays so far!
Yes, thanks. I'm now about to leave for Scotland for a couple of
weeks.
Without having checked, is it really true that "the majority
definition" has it the other way? What specifically constitutes
that majority?
Good question. Maybe I'm jumping to a conclusion here. The only
sources I've actually checked so far are Stock and Watson's textbook
and the R strucchange package.
Allin