[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
RE: st: RE: xi3 & postgr3 graph options
Not at all. I find it very difficult
to wrap my head around stories for all
but simple interactions. (Just saying e.g.
that "race, age and sex interact" is a
description, not an explanation, for example.)
Some people are lucky enough to work in fields
in which there is relevant theory.
This is all a matter of statistical style --
and statistical naivete (add accents there
to taste) -- and nothing to
do with e.g. programming experience.
More importantly, you are touching on
a perennial issue, which is why and which
user-written programs should be adopted
as part of official Stata. The criteria
that a user-written program are (a) well done
and (b) key to some users are
necessary but not sufficient...
> > Although the help file seems not to mention it, -xi3- is by
> > Michael Mitchell and Phil Ender of UCLA, and -postgr3-
> > is by Michael Mitchell.
> > Neither appears to be a member of Statalist at present,
> > so I'd recommend them emailing them directly for support.
> > I've not used either program myself, being among the very
> > simple-minded souls whose needs are met by -xi-.
> > Nick
> > email@example.com
> Thanks. You are missing out, then. I emailed Dr. Mitchell directly as
> well. Compared to R, Stata modeling statements make you work
> too hard.
> For simpler tasks with only two-way interactions, xi is useful, but
> even here it would be improved by access to the various
> coding schemes
> available. Postgr3 is like your anovaplot program but works with most
> Stata estimators. When coupled with xi3, which allows 3rd order
> interactions and all the various regression coding schemes (effect,
> helmert, adjacent, etc.), postgr3 makes graphical interpretation more
> rapid, more consistent, and for me, more pleasurable. My
> guess is that
> your "simple-mindedness" is made possible by your deeper
> experience and coding skill.
* For searches and help try: