[S] Special log-linear models for matched-pairs data

Robert Balshaw (rbalshaw@setosa.uwaterloo.ca)
Sat, 9 May 1998 18:00:05 -0400 (EDT)


I am looking at some particular log-linear models based on matched pairs,
such as are discussed in Chapter 10 of Agresti's "Categorical Data
Analysis".

I would like to be able to construct tests of hypotheses such as
conditional symmetry and quasi-symmetry.

These models can be expressed as restrictions on the parameters of the
completely specified model for the frequencies in a contingency table, but
I don't know how to fit these models in S-Plus. I'm hoping that someone
has already tackled this problem.

As an example, consider the I x J x K x K table on factors A, B, C1 and
C2. Factors C1 and C2 represent a K x K cross-tabulation of
subjects respondents to a K-category question at two times.
How would one test (in S-Plus) whether the C1 by C2 table is symmetric,
after conditioning on A and B, for example? Or, is there apparent
"trend" in the response from time 1 to time 2.

Any suggestions would be most welcome. I will summarize responses to the
list.

Thanks,

Rob

-- Robert Balshaw Statistics & Actuarial Science --
-- rbalshaw@setosa.uwaterloo.ca University of Waterloo --

-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news