Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: specifying offset in logistic regression


From   Maarten buis <maartenbuis@yahoo.co.uk>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: specifying offset in logistic regression
Date   Sat, 11 Jul 2009 20:52:13 +0000 (GMT)

--- On Sat, 11/7/09, Kathryn L Adeney wrote:
> I am trying to account for time-at-risk ("risktime") as an
> offset variable in a logistic regression model.  As far
> as I understand, it should not matter whether I use the
> option "exposure(risktime)" or "offset(lnrisktime)", where
> lnrisktime = ln(risktime), in the glm.  However, I get
> two different results depending on how I specify the
> option.  Is this a glitch that anyone else has come
> across? 

If you created lnrisktime as a float variable (the default)
you will get slightly different results differing in the
last one or two digits. These differences will disappear
if you create lnrisktime as a double. If this is the 
problem that you came across then this is not a glitch 
but a sign that Stata performs internal computations at
high precision. Below is an example where these two work
exactly as expected once lnpyears was created as a double.

*------------ begin example --------------------
webuse dollhill3, clear
tab agecat, gen(agecat)
gen double lnpyears = ln(pyears)
glm deaths smokes agecat1 agecat2 agecat3, ///
    exposure(pyears) link(log) family(poisson)
glm deaths smokes agecat1 agecat2 agecat3, ///
    offset(lnpyears) link(log) family(poisson)
*----------------- end example ------------------

Hope this helps,
Maarten

-----------------------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany

http://home.fsw.vu.nl/m.buis/
-----------------------------------------


      

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index