Hi Bioconductor Team!

Can regularized log/rld transformed values of 16S microbial taxon abundances be treated as independent of each other? Or are the resulting values for each taxon not independent because the transformation involves using the experiment-wide trend of variance over the mean?

Is it statistically appropriate to use rlog transformed microbial abundances in regression against an explanatory variable?

I'd be grateful for any suggestions or resources you know of!

I prefer the VST. It’s like a log with pseudocount but adapted to the mean variance curve. Its a fixed transformation so I wouldn’t worry about sample independence after. Think of it similar to log transformation.