Re: [Gretl-users] Chow/QLR
On Wed, 23 Sep 2009, Kehl Dániel wrote:
I have a question concerning Chow test and a cross-sectional
database. Is it possible to test a series of observations
somehow? Isn't it possible to do a test like QLR for time
series? The observations have to be in ascending/descending
order, and one could say the largest F value of the Chow test is
at observation X. Does this make sense at all?
Whether a QLR-type test makes sense for cross-sectional data is
really up to you to decide; I suppose it _might_ make sense in
some contexts (with the data ordered in certain ways), though not
in general.
Note that for the regular Chow test in gretl you have the --dummy
option to the "chow" command, designed for use with cross
sectional data. If you really reckon a QLR test makes sense this
is not something that gretl supports by default for cross
sectional data (too weird!) but you can easily fool gretl into
doing the test for you. Order the dataset as you wish
dataset sortby <whatever>
then pretend it's a time series
setobs 1 1 --time-series
and go ahead!
Allin Cottrell