[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Incidence rate ratio

From   Maarten buis <>
Subject   Re: st: Incidence rate ratio
Date   Thu, 6 Mar 2008 15:20:01 +0000 (GMT)

--- Mohammed El Faramawi <> wrote:
> I am trying to calculate the incidence rate ratio(IRR)
>  for a categorical variable ( Mortality) which  has
> two categories only ( dead or alive) in a probability
> weighted sample. I am also trying to adjust for
> covariates. I know that poisson regression can be used
> to calculate  (IRR) and adjust for covariates given
> that the outcome is  a count for example the number of
>  deaths, number stroke attacks, etc . My question is
> what regression should I use  in such a situation i.e
> outcome which is not count.Can I still use poisson
> regression which I doubt it?? Thank you

--- wangxin wrote:
> so, your dependent variable is dead or alive? that is
> a binary one. you may use Logistic regression.

-logit- will give you odds ratios instead incidence 
rate ratios (or risk ratios as they are often called 
in case of a binary dependent variable). There are 
people who think that odds ratios are too dificult 
to understand, for that reason prefer risk ratios or 
even risk differences. I disagree. For various views 
on this you can follow the thread starting with:

One problem with risk ratios is that it implies a model 
which can result in  predicted probabilities larger
than 1 if you have a continuous explanatory variable or 
multiple discrete explanatory variables. If you are willing
to live with that, than you can use -poisson- to estimate
this model. Alternatively, you can use -glm- with the 
-family(binomial) link(log)- options. The advantage of the
latter command is that it warns you when your model results
in predicted probabilities larger than 1.

Hope this helps,

Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434

+31 20 5986715

Sent from Yahoo! Mail.
The World's Favourite Email
*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index