Date:         Thu, 3 Jul 2003 13:46:25 -0400
Reply-To:     Jay Weedon <jweedon@EARTHLINK.NET>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Jay Weedon <jweedon@EARTHLINK.NET>
Organization: http://extra.newsguy.com
Subject:      Re: Reference Coding in Logistic Regression
Content-Type: text/plain; charset=us-ascii

On 3 Jul 03 13:46:16 GMT, pluo@DRAFTNET.COM (Luo, Peter) wrote:

>I have a variable A that has two values: M and F, and another variable B
>which can take 1, 2, or 3. I used reference coding to transform A (reference
>category is F) and B (reference category is 3), and then run a logit
>regression with A*B as the interaction term. I got the following estimates
>for the interaction term
>
>
>
>A        M          (parameter estimate)
>
>B        1           (parameter estimate)
>
>B        2           (parameter estimate)
>
>A*B   M 1       (parameter estimate)
>
>A*B   M 2       (parameter estimate)
>
>
>
>My question is: what M 1 was estimating? Was it M1 vs F1, or M1 vs. M3?

It estimates the log OR for M vs F at B=3.

JW