Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: Re: st: Poisson regression


From   [email protected]
To   [email protected]
Subject   Re: Re: st: Poisson regression
Date   Mon, 01 Aug 2005 08:54:11 +1000

Thank you Svend for the answer.
So I shouldn't worry even if it is as small as 0.02%?
Azi



> Svend Juul <[email protected]> wrote:
> 
> Azi wrote:
> 
> I am looking at the predictors of standardised mortality
> 
> ratio (SMR) for my cohort. I have two datasets:
> 
> -the first one contains death; person-years of observation
> 
> and different variable which i want to check as predictors
> 
> of smr.
> 
> -the second dataset contains the rate of mortality 
> 
> (adjusted for sex and age) for the reference population.
> 
>  
> 
> After stset and stsplit in the first dataset, i merged it
> 
> with the second one and calculated the expected number of
> 
> death (e) for my cohort. In order to look at the 
> 
> predictors of smr i used the following syntax:
> 
>  
> 
> xi: poisson _d  i.psy  i.cob ......., exposure(e) irr. 
> 
>  
> 
> I wonder whether the selection of "e" as the exposure is 
> 
> correct when my dependent variable is smr?
> 
>  
> 
> When i fit different models, pseudo R^2 is very small 
> 
> (around 0.02) in all of them. This happens even when I 
> 
> include variables which influence smr. For all models use of
> 
> poisgof gives a good results.
> 
> ------------------------------------
> 
>  
> 
> From what you tell us, you did it right, provided the second
> 
> dataset contains reference mortality rates for each sex and 
> 
> age group (not "adjusted" for sex and age). Just one phrase:
> 
> I wouldn't call smr the dependent variable; it is a measure
> 
> of association or contrast (the dependent variable is death).
> 
>  
> 
> Don't worry about the small pseudo R^2. The interpretation
> 
> is dubious for dichotomous outcomes; in my general
> 
> understanding a high R^2 would mean that we were able, from
> 
> the model, to predict when each individual would die - and
> 
> we are hardly that clever yet. 
> 
>  
> 
> Good luck,
> 
> Svend
> 
>  
> 
> ________________________________________________________ 
>  
> Svend Juul
> Institut for Folkesundhed, Afdeling for Epidemiologi
> (Institute of Public Health, Department of Epidemiology)
> Vennelyst Boulevard 6 
> DK-8000 Aarhus C,  Denmark 
> Phone, work:  +45 8942 6090 
> Phone, home:  +45 8693 7796 
> Fax:          +45 8613 1580 
> E-mail:       [email protected] 
> _________________________________________________________ 
> 
> *
> *   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/
*
*   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/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index