I am interested in a disease prevalence in each group (total 4 groups). To determine statistical differences in proportions among these groups, I would like to know how many people I need in each group. Can I use "proc power multreg", Or proc power twosamplefreq (Bonferroni correction, alpha=0.0083). If I use proc power twosamplefreq, do I need to set three different odds ratio in reference to control and alpha adjustment (0.0083) with Bonferroni correction (6 comparison, alpha 0.05/6=0.0083). Thank you.
proc power;
multreg
model = fixed
nfullpredictors = xx
ntestpredictors = 3
rsquarefull = xx
rsquarediff = xx
ntotal = .
power = 0.8;
run;
proc power;
two samplefreq test=lrchi
oddratio=1.5 2 2.5
ref proportion=0.5
npergroup=.
alpha=0.05 0.0083
power=0.8;
run;
