[Gretl-users] Re: "Data Error" from kpss_test/adf_test C/C++ api
Am 14.05.2021 um 18:23 schrieb riles(a)triton.net:
has it just work for all the other functions because they don't
require things to be "Real"? or check that they are "Real"?
Econometrically/statistically It depends on the context. It's not absurd
to run a difftest on a pair of variables where one is constant. Although
it might be debatable whether the built-in constant term with all ones
should be allowed; the alternative could be to require the user to
create a wanted constant series.
For other tests like KPSS some minmal variation is really needed to make
it work numerically, so banning the constant term makes sense.
Don't know about whether "all the other functions" really work with the
constant.
cheers
sven