01-21-2015 05:05 PM
I am trying to calculate the sample size for a paired freq test using mcnemar. I have estimated that my discordant proportions are p01=.0012 and p10=.0138. My hypotheses are:
Ho: p01>=p10 OR p01-p10 >= 0
Ha: p01 < p10 OR p01-p10 < 0
Since I only care about the effect size if p01 > p10, I thought I should change the values of p01 and p10 to both = (0.0012+0.0138)/2 = 0.0075 Then, my proportion difference is 0; my reference proportion is 0.0075
However, when I try this in proc power, I end up not being able to test the effect size = 0 and receive the error message:
WARNING: The magnitude of the effect is zero in 4 scenarios in the output table.
My code is posted below. Can someone help me to understand how to test this hypothesis?
power=.8 .85 .9 .95
discpropdiff = 0
refproportion = 0.0075