Date: Mon, 23 Jan 2006 09:49:26 -0500
Reply-To: Peter Flom <flom@NDRI.ORG>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Peter Flom <flom@NDRI.ORG>
Subject: Re: QR:Logistic regression (1 values are much more less than the
0 's)
In-Reply-To: <20060123144020.93824.qmail@web26404.mail.ukl.yahoo.com>
Content-Type: text/plain; charset=US-ASCII
>>> "adel F." <adel_tangi@yahoo.fr> 1/23/2006 9:40 am >>> wrote
<<<
Thanks,
My total sample is 7350, and I have got 16 categorical variables,
most of them are binary (2 values) others with 3 to 4 categories, I am
interested for the moment to the main effect.
I will be interested to some interactions
What will be the problems if I consider the logistic regression in
the situation that I have described?
Any comments are very welcome
>>>
and earlier
<<<
I have a simple question, I want to model a binary variable using
proc logistic, my depedent variable has 5% values of 1 and 95% values
of
0.
>>>
So 5% of 7350 = about 350, and you have 16 IVs, so, at least without
interactions,
you should be OK (there's a rule of thumb that you want 10 cases in
your smaller outcome
for each IV).
I will let people with more knowledge comment on whether it's good
rule of thumb.
Of course, if some of the categories on your IVs are very small, that
might cause problems.
Peter
Peter L. Flom, PhD
Assistant Director, Statistics and Data Analysis Core
Center for Drug Use and HIV Research
National Development and Research Institutes
71 W. 23rd St
http://cduhr.ndri.org
www.peterflom.com
New York, NY 10010
(212) 845-4485 (voice)
(917) 438-0894 (fax)
| 
