SAMPLE SIZE for a cross over AB/BA with binary variable

[COMPAREMISSING]

I have a cross over AB/BA with binary variable. I have 16 subjects who take 2 treatments in 2 periods. the 16 are diveded in 2 groups which follow 2 diffrerent sequences:AB and BA

I have to analyze it as:

PROC GENMOD DATA= sample DESCENDING;
CLASS TREATMENT PERIOD  SUBID;
MODEL Y = TREATMENT PERIOD /LINK=LOGIT DIST=BIN;
REPEATED SUBJECT=SUBID ;
RUN;

how i can calulate the power?

Thanks a lot,

Barbara

[SteveDenham]

Simulation is your best bet.  However, with transformations (logit) you might be able to use PROC GLMPOWER.  You will need good estimates of the correlation over time to get an accurate estimate.  If you have SAS/STAT14.1, look at Example 48.3 Repeated Measures ANOVA as a starting place.

Steve Denham