Re: st: Messages not appearing; Subject: Specifying interactionsin loglinear models

[Joan Holand <Joan.Holand@gmx.net>]

Dear all,

I'm running (Stata 9.2, Windows XP) a glm to specify a loglinear model
with 3 variables. To obtain the interaction effects I'm using the
xi3-extension with additional coding schemes:

xi3: glm depvar e.var1*e.var2*e.var3, fam(poi)

I'd like to get the coefficients for *all* interactions (without
omitting any category). By the means of xi3 (@ operator) one can get
only the 2-way-interactions.

My questions:
1. Is there any way for getting three (or more)-way-interaction terms
without omitting any category and disposing the additional coding (say
effect coding) at the same time?

2. In the standard output I only get the coefficients of the additive
model (the logarithm of the respective effects). (How) Can I display the
coefficients of the multiplicative model (not logarithmized), too?

Many thanks in advance
Joan

P.S. This is my fifth try to post this message since Saturday midday (I
used two diffrent e-mail accounts!) ... I haven't found it in the
statalist archive up to now. Sorry for posting it as an answear, but
my question is getting urgent: Maybe this time I'll be more lucky and it
will pass through the statalist spam filter :-)


Marcello Pagano schrieb:
> Dear Listers,
>
> Some odd behaviour in what appears and what does not appear on the
> list lately may be traceable to our spam detector we have operating
> locally. It daily captures about 100,000 letters.  For me personally
> it stops about a 100 letters that have mostly to do with offers to
> enlarge part of my anatomy, so I am grateful the program is in place!
> Of course, the program makes decisions based on some judgment calls,
> and as a result it sometimes does things we do not expect.  So if
> your posting to Statalist does not appear within a short time
> (usually, a matter of minutes, not hours) ask yourself the usual
> questions: (1) Am I still subscribed? (sometimes you get unsubscribed
> because letters to you get bounced back to us, or because you go on
> vacation and we get those pesky "Ta ta, I am on vacation!"
> messages.), (2) Is the message too long? (Not usually the case.), (3)
> Am I sending a formatted message? (Usually the case. Remember,
> Statalist requires "Plain Text", and does not like (a common 
> courtesy, please get rid of) those signature cards that get attached
> to messages by some.), and (4) Am I mailing from another account?
>
> And now you can add another reason: possibly the spam detector
> swallowed up my message.  Currently, the only way around that is to
> wait a decent amount of time.  Check the archives to see if your
> message appears there.  If not, resubmit the message, but slightly
> transformed. Possibly change an URL--e.g. put a > in front of it so
> that it looks like text and not an URL. Use your imagination. Don't
> talk about enlargements? Anyway, it should go without saying that you
> should check,
>
> http://stata.com/support/faqs/res/statalist.html
>
> m.p.
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/