Date: Wed, 10 Apr 2002 15:20:41 -0400 "Beckman, Anthony" "SPSSX(r) Discussion" "Beckman, Anthony" Re: Gender codes To: Jeremy Miles text/plain

Thanks for the terriffic references!

-----Original Message----- From: Jeremy Miles [mailto:jnvm1@york.ac.uk] Sent: Wednesday, April 10, 2002 12:46 PM To: SPSSX-L@LISTSERV.UGA.EDU Subject: Re: Gender codes

At 11:11 10/04/02 -0400, Jessica Kenty wrote: >I was interested in learning the answer to Anthony's question as well. >Thanks Jeremy for your response. I have a followup question about this >though. Why wouldn't you just create a dummy for this variable (i.e. >0=male; 1=female)? Do we lose information when we do this? Also, what >about when we have a categorical variable with 3 or more values - can you >still do effect coding this way? > >Just curious...

They differ in terms of the interpretation of both the intercept and the effects.

Dummy coding (note that not everyone uses the same terminology all of the time), there is a control group identified, this is labelled as a zero, the other groups are labelled as 1.

So if group is the original variable, and we wich to use group 1 as the reference group:

Group Dummy_2 Dummy_3 1 0 0 2 1 0 3 0 1

The intercept is the mean score for group 1, and the betas are the size of the difference between the mean for each group and group 1. The t-tests test the difference between each group and the mean.

In effect coding:

Group Dummy_2 Dummy_3 1 -1 -1 2 1 0 3 0 1

Here, group 1 is still the reference group, but the betas compare the means for each group with the overall mean for all groups. This is useful when you don't have a natural control group.

This is covered in most books on regression (including mine :) Miles and Shevlin, 2001, applying regression and correlation, but is also in Hutcheson and Sofreniou, 2000, the multivariate social scientist, Cohen and Cohen, 1982, Applied multiple regression/correlation analysis, and Pedhazur, 1998?, multiple regression in behavioural research, to name some of my favourites.

Jeremy

Jeremy Miles, Lecturer in Biostatistics, Department of Health Sciences, Alcuin COllege, University of York, YO10 5DD mobile: 07941 228 018 mailto:jnvm1@york.ac.uk http://www-users.york.ac.uk/~jnvm1 [working at home] http://www.jeremymiles.co.uk

Back to: Top of message | Previous page | Main SPSSX-L page