st: Assistance on variable selection problem

["Lachenbruch, Peter" <Peter.Lachenbruch@oregonstate.edu>]

Dear Lachenbruch, Peter <Peter.Lachenbruch@oregonstate.edu>,

You sent email to Statalist <statalist@hsphsun2.harvard.edu> and Majordomo,
the Statalist software, bounced it.  The email you sent is attached below, but
the most likely reason for the bounce is:

Admin request: /^subject:\s*help/i    

Majordomo might also have bounced it for any of the following reasons:

    1.  You are not subscribed to Statalist, or not subscribed 
        under the email address from which you sent the email.
        Please read section 2.2 of the Statalist FAQ for advice
        on this issue:

          http://www.stata.com/support/faqs/res/statalist.html#toask

    2.  You sent the email in HTML format.  Please read section 2.2 of
        the Statalist FAQ for advice on this:

          http://www.stata.com/support/faqs/res/statalist.html#toask

    3.  You sent a subscription request to the wrong address.  To
        subscribe, send an email to majordomo@hsphsun2.harvard.edu
        with "subscribe statalist" in the body of the message, not
        in the subject line.  The subject line does not matter.  Please
        use plain text when communicating with the majordomo software.
        
    4.  You sent an unsubscription request to the wrong address.  To
        unsubscribe, send an email to majordomo@hsphsun2.harvard.edu with
        "unsubscribe statalist" in the body of the message, not
        in the subject line.  The subject line does not matter.

    5.  Your email contained a word in its first few lines which majordomo
        thought was an administrative request intended for majordomo,
        such as "help", "subscribe", or "unsubscribe".  Avoid these
        words in the first few lines of your email.

There are other possibilities.  See the Statalist FAQ, and especially
section 2.2, at http://www.stata.com/support/faqs/res/statalist.html

Regards, 

Statalist

P.S.  Why is this email coming from stata.com?  I thought Statalist
      was independently operated!

      Answer:  It is.  Marcello Pagano, Statalist's moderator,
      posted the following on Statalist on March 11, 2010:

        As Statalist has grown, so have the number of problem emails
        sent to Statalist that don't appear on the list for one reason
        or another.  People try to post on Statalist who are not
        subscribers. Subscribers sometimes send valid email, but in
        HTML format. Subscription and unsubscription requests are
        sometimes sent to the list rather than to majordomo. 
        In general there are 50 to 100 problem emails per day! Up 
        until now, we've been handling them by hand.

        This cannot continue. There is an alternative. I've talked to
        StataCorp and we can forward all the bounced messages to them
        and they can set up an automatic response system to send the
        email back to the sender along with some indication of the
        problem.

        I want to take advantage of their offer to help, but I do not
        want anyone to think that StataCorp's involvement is anything
        more than helping with this technical issue.  All emails that
        get posted the list will still be posted to the list without
        ever going to StataCorp and not subject to any censoring.
        All StataCorp will receive are the emails Statalist would have
        bounced anyway, and all they will be doing is using their
        software to analyze those emails and send back an automated
        response to the sender.

        I assume everyone is okay with that. If someone is not,
        I'm willing to forward the bounces to them and let them
        handle them :-)

Here is the email you sent to statalist@hsphsun2.havard.edu:

Received: from guardian.sph.harvard.edu (guardian.sph.harvard.edu [134.174.190.3])
	by hsphsun2.harvard.edu (8.11.7p1+Sun/8.11.7) with ESMTP id p5AGeJA21385
I reread the bounce I got and apparently it was the word "help" in the title.  Apologies if this appears twice.

This is not especially a Stata question, but it is driven by an analysis issue...

A student is trying to analyze data from a national survey (no weights needed).  She has 26 variables plus 10 years of data.  There are about 1,000,000 observations.  With this many observations, everything is significantly different from 0.  She's using mlogit (predicting medical care expenses), so she'd like to cut down the number of 'important' predictors.  I have thought of several options: backward stepwise  (not available with mlogit); look at effect size and insist it be larger than 0.05 - again not available since there are four categories of the response variable; use a Bonferroni inequality on the coefficients and insist on a low p-value to begin with - e.g. try for a size of 0.01 adjusting for 25 tests, so p must be less than 0.0004.  The issue seems to be the huge sample size pushing everything to significance.
Does anybody have any ideas?

Tony

Peter A. Lachenbruch
Department of Public Health
Oregon State University
Corvallis, OR 97330
Phone: 541-737-3832
FAX: 541-737-4001




*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/