transformation for plots about ancombc HOT 2 CLOSED

Hi Anny,

Thanks for your interest in ANCOM and ANCOMBC! Please kindly find below my responses:

I'm wondering what is the best way to plot changes in abundances of DA features identified with ANCOM-BC.

As for the primary analysis, visualizations using either beta values (unstandardized effect size) or W statistics (standardized effect size) would be a good way to show changes in absolute abundances. A positive value indicates an increase in absolute abundances as compared to the reference level; a negative value, on the other hand, indicates a decrease in absolute abundances. This vignette shows an example of using waterfall plots.

I used to plot clr-transformed counts on heatmaps when I was using ANCOM but now that I switched to ANCOM-BC I get very conflicting results.

Indeed, it happens sometimes that the clr-transformed values and ANCOMBC W statistics give a contradictory answer, which is basically because clr transformation relies on the geometric mean of observed abundances as a reference. For example, if taxon X is more abundant in treatment A than treatment B; however, if the geometric mean in treatment A is even larger than that of treatment B, the clr-transformed values of taxon X will show an opposite direction. I would recommend using W statistics as the input values for your heatmap so the interpretations will be more consistent.

Hope it helps :)
Huang

from ancombc.

Thank Huang!
Very helpful and makes complete sense.
All the best,
Anny

from ancombc.