st: Re: Logistic regression and confidence intervals (a little long)

[Dario Consonni <Dario.Consonni@unimi.it>]

Title: Logistic regression and confidence intervals (a little long)

Assuming you already tested the interaction term type*dumclass3 and it was not significant, use -lincom- to get OR and CIs for the OR when both type and dumclass3 are present:

 

   lincom type+dumclass3

 

Ciao, Dario

 

 

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

On August 29, 2002 10:28 AM T.Petitti@unicampus.it wrote:

Dear STATAusers,
Suppose I perform a logistic regression on this dataset:

Variable        Data
-----------------------------------
sym             binomial               
type            binomial
dumclass3       binomial

So the command is:

logistic sym type pH

Suppose I've this result:
------------------------------------------------------------------------------------
Logit estimates                                   Number of obs   =       4848
                                                  LR chi2(2)      =     297.12
                                                  Prob > chi2     =     0.0000
Log likelihood = -752.84441                       Pseudo R2       =     0.1648

------------------------------------------------------------------------------
         sym | Odds Ratio   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
        type |   10.98154   1.644385    16.00   0.000     8.188485    14.72729
   dumclass3 |    1.63897   .2841108     2.85   0.004     1.166857    2.302101
------------------------------------------------------------------------------
                                                               
------------------------------------------------------------------------------------
Now ... it's clear that when dumclass3=0 then OR being type=1 and sym=1 is 10.98,
and when dumclass3=1 then OR being type=1 and sym=1 is 10.98*1.63=17.89

The question is: which is the CI for the OR 17.89? Is it 8.1*1.1=9.4
 and 14.7*2.3=33.8 (i.e. the product of the two CI)?
    
Thanks for the anwers you'll send me.

     Dr. Tommasangelo PETITTI
---------------------------------------
         Medical Directorate
 University Hospital Campus Biomedico
           +39-6-22541521