topic Analysis of 2 x n contingency tables with matched pairs of subject in Statistical Procedures

Analysis of 2 x n contingency tables with matched pairs of subject

AnalytX — Thu, 19 Jul 2012 07:37:02 GMT

Hello,

I use McNemar test via PROC FREQ to analyse 2 x 2 contingency tables with matched pairs of subject. What statistical test should I use (and corresponding PROC) to analyze 2 x n contingency tables with matched pairs of subject.

Thanks in advance for you suggestions;

Regards,

Re: Analysis of 2 x n contingency tables with matched pairs of subject

SteveDenham — Thu, 19 Jul 2012 12:07:47 GMT

I may be a little unclear on this, but I would try using a generalized linear model that captured the pairing design. Perhaps something like:

proc glimmix;

class group pair;

model nlevel = group / solution dist=multinomial;

random int / subject=pair;

run;

Without knowing exactly what the data look like, this is as close as I can come up with for paired multinomial data.

Steve Denham

Re: Analysis of 2 x n contingency tables with matched pairs of subject

AnalytX — Fri, 20 Jul 2012 07:45:43 GMT

Dear Steve,

Thanks a lot for your answer which sounds very good.

I have performed some tests and I obtained the results I waited for.

Have a good day.

Regards