I'm testing some methods for estimating correlation in compositional data. As part of this process, I'm using the following approach:
- simulate un-normalized features $Y$
- normalize features for each sample to get compositional data $X$
- attempt to estimate correlation of $Y$ from $X$
- calculate false positive/negative rates
I'm exploring using a permutation test to identify significant correlations (my positives). Should I be permuting my data before or after normalizing it (step 2)?