Date:         Mon, 1 Sep 2003 10:13:31 +0200
Reply-To:     Asesoría Bioestadística
              <bioestadistica@eresmas.net>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Asesoría Bioestadística
              <bioestadistica@eresmas.net>
Subject:      Re: rules for presenting results from logistic regression
Content-Type: text/plain; charset=us-ascii

Hi Frank


> a simple question: Are there any rules how to present results from
> logistic regression published on the web?

I don't know of any Web resource for taht, but this could be helpful:

Thomas A. Lang, Michelle Secic (1997): "How to Report Statistics in
Medicine". ACP Series, Philadelphia. Chapter 7 (Reporting Regression
Analyses), p 120, table 7.2.

The suggest the following information to be given (best viewed with
fixed width font):

Variable       Coeff (beta)   SE(beta)  Wald Chi^2  P-value   OR
95%CI
------------------------------------------------------------------------------

Var#1 (units)     0.44          0.11      16.00     <0.001   1.55 1.25
to 1.93
Var#2 (units) ...........
Var#3 (coding)......


------------------------------------------------------------------------------

If the IV variables are quantitative, indicate always the units (cm,
years, pounds...).
If they are qualitative, indicate the categories, as well as the
reference cell (Non smokers and Ex-smokers vs Smokers, Male/Female...).

Nonetheless, I normally use an abreviated form (not my invention, it was
indicated to me by several good epidemiologists & statisticians):

Variable      LR test     P-value          OR          95%CI

The use of LR test instead of Wald test is recommended by
Hosmer&Lemeshow in their book ("Applied Logistic Regression"), due to
the low power of Wald test.

HTH

Marta Garcia-Granero