st: RE: marginal coding of dummy varaible

 From "Maarten Buis" To Subject st: RE: marginal coding of dummy varaible Date Wed, 26 Apr 2006 17:14:55 +0200

```Valérie Jooste:
In this coding scheme all parameters (including the one left out)
add up to zero. If you have 5 groups, you will get 4 parameters
and the fifth parameter is
-(_b[group2]+_b[group3]+_b[group4]+_b[group5])
HTH,
Maarten

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

Buitenveldertselaan 3 (Metropolitan), room Z214

+31 20 5986715

http://home.fsw.vu.nl/m.buis/
-----------------------------------------

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu]On Behalf Of Valérie Jooste
Sent: woensdag 26 april 2006 17:04
To: statalist@hsphsun2.harvard.edu
Subject: st: marginal coding of dummy varaible

I am trying to perform a logistic regression without defining a reference
group for the independent discrete variables:
I am intending to determine the ratio of the odds of each of the J levels of
a discrete variable to the average odds (geometric mean of all levels). For
that matter, instead of using the classical reference coding, I followed the
"marginal method". I created dummy variables using the marginal coding:
It means that, instead of coding the jth dummy variable 1 if the discrete
variable is at the jth level and 0 if the discrete variable is at any other
level, I coded the jth dummy variable -1 for the first level, 1 for the jth
level and 0 for the others.
This works beautifully and avoids having a group as reference but brings
another (big) problem: the 1st level doesn't appear at all in results (not
even as reference...).
How can I estimate a coefficient for level 1?

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```