Date: Sat, 26 May 2007 13:17:54 -0700
Reply-To: "Dennis G. Fisher, Ph.D." <dfisher@CSULB.EDU>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: "Dennis G. Fisher, Ph.D." <dfisher@CSULB.EDU>
Subject: Re: logistic regression wierdness
In-Reply-To: <23813195.1180207896293.JavaMail.root@mswamui-blood.atl.sa.earthlink.net>
Content-Type: text/plain; charset=UTF-8; format=flowed
Peter,
MV is ever use of both Methamphetamine and Viagra. The data actually
make sense. The question is what to do about it?
Thanks.
Dennis
Peter Flom wrote:
> Hi Dennis
>
> well, unless I am reading this wrong (it got formatted oddly) you only have 2 people in one of the cells....so, things are going to be weird! One could say that ANY model is overfit, because a single noise variable will suffice to put 1 person in the right cell
>
> What is MV?
>
> Peter
>
> -----Original Message-----
>
>> From: "Dennis G. Fisher, Ph.D." <dfisher@CSULB.EDU>
>> Sent: May 26, 2007 1:58 PM
>> To: SAS-L@LISTSERV.UGA.EDU
>> Subject: logistic regression wierdness
>>
>> We are trying to construct a logistic regression model. When we put
>> ever use of marijuana in the model, we get an odds ratio of 5.414 and
>> the model seems to fit. We still get another 10 variables that seem to
>> be OK. Usually when this kind of thing happens, I figure it is either
>> dirty data or we coded something wrong, but this time I think it is
>> real. The 2 x 2 table is
>> Marijuana
>> No Yes
>>
>> MV No 194 443
>>
>> Yes 2 165
>>
>> This is almost seems like a a quasi complete separation, but it really
>> is not I don't think. I am going to feel really strange trying to
>> present a model with an OR this big, but I cannot figure out why I
>> should take it out of the model. What would you do? Thanks.
>>
>> Dennis Fisher, Ph.D.
>> Center for Behavioral Research and Services
>> California State University, Long Beach
>>
>
>
>
| 
