I would have done it like this. Generated a column of Ratio (of Test/Reference) and then PROC MEANS with CLM option get the confidence limits. Since I have not seen the data, I might have severely underplayed the complexity you are facing. If the data is not sensitive you can paste a few rows for us to understAnd the problem better. As far as log transformation is there, I generally do it if the range of numbers is very large (say from 1 to 1Billion) or very small from (0.000000001 to 1). If you do log transformation the range will come out to be 0 to 9 in the first case and -9 to 0. The benefit is your X and/or Y axis is not stretched so that cognitive load on the brain is less. So it depends, in my opinion, what is your range. As far as Norm for Acne is concerned (for log transformation) Google might know the answer. You just might need to patiently talk to it. Try Pubmed (NCBI) papers. See if there is a precedence. All the best.
... View more