Dear Statalisters:
I would like to ask your help for analysis of associations in
three-way contingency table.
Three variables include a response variable and two control variables
of district and ethnicity. Data was collected through stratified
cluster sampling so that I need to use “svy” command or any kind of
adjustment of sampling weight. Because this research is to compare
relationships of three ethnic groups in two district, the sample sizes
of 3 ethnicity and 2 districts are controlled (about 110 samples per
one ethnic group so that 330 samples per district, and the total
sample size is about 660.) A response variable is socio-economic
status, education level or level of trust, i.e. categorical or ordinal
variable.
First I tried to use a log-liner model for analysis of contingency
table by glm family (poisson) link(logit). But I don’t know how to
incorporate sampling weight because dependent variable of log-linear
model is three-way table. If sampling weight is ignored,
ethnicity*district model is significant of course. Then, svy, subpop
( ): tabulate would be another option (for conditional independence
analysis). So my question is, can I use this svy tab with subpop as
alternative approach to log-linear model? Or is there any other
approach that I can take? (maybe one way is simply to use svy: logit
though)
The second question is what kind of test of independence can be used
for analysis of association between ordinal and nominal variables in
three-way table? There are many tests for ordinal-ordinal variable
associations, but not ordinal-nominal associations. The
Cochran-Mantel-Haenszel test is one test, but I cannot find the code
for CMH test especially for survey data.
I will truly appreciate your suggestion and advice.
Thank you very much.
Best regards,
Hisako
--
********************************
Hisako Kobayashi
[email protected]
Ph.D Candidate in Public Administration
School of Policy, Planning and Development
University of Southern California
********************************
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
