Re: [Gretl-devel] difftest
Am 29.11.2017 um 01:34 schrieb Allin Cottrell:
On Tue, 28 Nov 2017, Sven Schreiber wrote:
> 1) Can we have a --quiet or --silent switch to access only the
$test
> and $pvalue accessors in a script?
OK, you now have "diffest ... --quiet" in git.
Thanks!
> 3) Could the parametric t-test of differing means in (paired)
samples
> be subsumed under difftest? (Currently there is no direct scripting
> way of performing it, I think.)
lib/src/nonparam.c: have at it!
The idea was to use _existing_ code that is used in the test statistic
calculator ("2 means") and hook that up to difftest. I guess that would
be do_two_means_test in gui2/calculator.c.
Because AFAICS you cannot perform the two-sample t-test in hansl (or can
you?), and that would seem to address also this comment in calculator.c:
"/* FIXME : should we record a relevant command when a test is done
using dataset variables? (Two means, or two variances)
*/"
I'm quickly getting lost in the internal variable and struct
definitions, however (test_t etc.).
> Furthermore, for small samples n=5..8, gretl appears to
calculate... W+ and
> W-.
We're looking for a z-statistic, and that's NA for the given
sample
size. If anyone wishes to pursue tiny-sample statistics in this area,
that's fine by me.
Fair enough. It's just that my impression was that gretl already wanted
to go in that direction, because it is reporting some relevant critical
values and the W+/W- thingies.
thanks,
sven