PROC SURVEYFREQ

[jrohay]

Can PROC SURVEYFREQ be used to conduct a McNemar's test?  If not, can PROC SURVEYMEANS be used test whether the difference in the proportions is 0 (i.e., compute a difference score and get the T and PROBT stats)? If so, are there any references for such an analysis?

[ballardw]

Since McNemar's test doesn't apparenlty use variance in calculations do you need SurveryFreq? Proc Freq will do one.

[SteveDenham]

I'll support @ballardw's suggestion.  Here is an example from a survey, that uses PROC FREQ:

http://www.stat.purdue.edu/~tqin/system101/method/method_mcnemar_sas.htm

That isn't to say that there aren't complex survey methods for McNemar tests out there (google is your friend for this)--it's just that from the info given, PROC FREQ seems adequate.

Steve Denham