K-sample marginal homogeneity test

[Kirill Orlov]

What are your thoughts about how one can perform K-related-samples marginal homogeneity analysis in SPSS?

Marginal homogeneity test is "repeated-measures ANOVA" for nominal categorical response. SPSS has such test in Nonparametrics, but it is for 2 related samples only. But I want an omnibus result for 3+ samples at once. SPSS has, in K-related samples, Cochran's Q analysis - put it is restricted for dichotomous responce only, while I have >2 nominal responses.

What are your thoughts? Is there a procedure in SPSS to K-related-samples marginal homogeneity analysis? I.e. to compare margial distributions in a (hyper)cubic frequency table with all dimensions defined by the same 3+ categories (such as Yes, No, DontKnow; or whatever). For example, could Loglinear or GEE be used for such a task? Or do you know a user-written syntax somewhere?

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

[Bruce Weaver]

Hello Kirill.  First of all, I don't claim any great expertise in this topic. ;-)  But when I Googled <tests of marginal homogeneity>, I noticed that John Uebersax has some notes on the topic, and I've found in the past that is stuff is generally very good.  His notes include a section called "Loglinear, association and quasi-symmetry modeling".  I think it might give you some optimism about using that approach.  

   http://www.john-uebersax.com/stat/margin.htm

HTH.

Kirill Orlov wrote

What are your thoughts about how one can perform*K-related-samples
marginal homogeneity analysis* in SPSS?

Marginal homogeneity test is "repeated-measures ANOVA" for nominal
categorical response. SPSS has such test in Nonparametrics, but it is
for 2 related samples only. But I want an omnibus result for 3+ samples
at once. SPSS has, in K-related samples, Cochran's Q analysis - put it
is restricted for dichotomous responce only, while I have >2 nominal
responses.

What are your thoughts? Is there a procedure in SPSS to
K-related-samples marginal homogeneity analysis? I.e. to compare margial
distributions in a (hyper)cubic frequency table with all dimensions
defined by the same 3+ categories (such as Yes, No, DontKnow; or
whatever). For example, could Loglinear or GEE be used for such a task?
Or do you know a user-written syntax somewhere?


=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD