Additive log-ratio on counts vs proportions

[aimirza]

Hi!

I love the Additive Log-Ratio (ALR) transformation because it inherently accounts for differences in library size in microbiome data. Given a component Xi and a reference component G, the ALR is defined as ALR(Xi) = log(Xi / G). Since both Xi and G are proportional to the total library size, their ratio effectively cancels out this library size effect, eliminating the need for separate normalization.

However, I noticed that ALR in ANCOM is applied to proportions and not counts, why is that?