Date: Wed, 26 Nov 2003 12:31:54 0500
ReplyTo: "Burleson,Joseph A." <burleson@up.uchc.edu>
Sender: "SPSSX(r) Discussion" <SPSSXL@LISTSERV.UGA.EDU>
From: "Burleson,Joseph A." <burleson@up.uchc.edu>
Subject: Re: Testing differences between categorical variables
ContentType: text/plain; charset="usascii"
Katherine:
You might consider a 2 (A or not) X 2 (B or not) X 4 (ethnicity) log
linear analyses (not logistic regression), sometimes called multiway
frequency analysis. This is the more general formula under which
logistic regression is one type. Here, no one variable is the dependent
variable, and the main effects (departure from 1:1 ratio) of each
variable, as well as the 3 2way and 1 3way interaction can be
assessed. The 2way interaction between the A X B is probably well
understood already. Your interest would center around the A X Ethnicity,
B X Ethnicity, and the A X B X Ethnicity. A whole series of model
testing can be done.
There are many good references, but Tabachnick & Fidell (4th ed.)
(~2001) Using multivariate statistics, is a good primer & synopsis,
while Agresti (1990) Categorical data analysis, WileyInterscience, is
technical and complete. My favorite, however, is still Wickens. T. D.,
Multiway contingency tables analysis for the social sciences (1980),
Erlbaum. He has wonderful examples, and sets a more moderate pace.
Joe Burleson
Original Message
From: SPSSX(r) Discussion [mailto:SPSSXL@LISTSERV.UGA.EDU] On Behalf Of
Katherine
Sent: Wednesday, November 26, 2003 3:52 AM
To: SPSSXL@LISTSERV.UGA.EDU
Subject: Testing differences between categorical variables
Dear List members,
I am trying to run a test which measures differences between 2 groups of
categorical variables. For example, I would like to find out the
differences of four different blood type (A,B, AB,O) between four races
(White, Chinese, Malay, Indian).
Is there any statistical method to use in such situation besides testing
of column proportions?
Thanks so much in advance.
Regards,
Katherine
