|Date: ||Mon, 21 May 2001 13:04:53 -0400|
|Reply-To: ||Jay Weedon <jweedon@EARTHLINK.NET>|
|Sender: ||"SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>|
|From: ||Jay Weedon <jweedon@EARTHLINK.NET>|
|Subject: ||Re: Proc logistic with 2% of event|
|Content-Type: ||text/plain; charset=us-ascii|
There's no particular reason why logistic regression could not be an
adequate model. The issue is not whether the analysis can be actually
DONE - it's how well the data FIT that type of model. When the
probability of the event is very small, Poisson models often fit
better than logistic models, that's all. Try both & see which fits
On 15 May 01 23:01:44 GMT, bendiabare@NETSCAPE.NET (barrere Bendia)
>I have data such that 2% of the event "dease= "1"
> and 98% event "dease= "0" then proportion is p=2% and 1-p=98%.
>I would make logistic regression with some Indep Vari: Var1, Var2, ...
>Some variables are categorical other are continous.
>Some people say that I can't do logistic regression because the proportion p=2% is too small. But when I do it, It seems all things OK. There's convergence,and..
>CAN SOME ONE give any comment or justification ...?
>Thank in advance.
>Get your own FREE, personal Netscape Webmail account today at http://webmail.netscape.com/