first time here and looking for help with two questions regarding songbird (thanks for the tool btw):

my q-score of both the null and baseline model are near zero and I tried quite a lot of adjustments to increase it. However, as my samples are all REALLY similar to each other, I assume this could be the reason?

I tried displaying the differentials in R using pheatmap and was just wondering if displaying the pure differentials is sufficient (I still cant really grasp the concept of intercept, even after the fairly well written blog post by one of the authors?

The Q2 score reported can be problematic - there doesn't seem to be a resolution getting null-hypothesis test for multinomial regression. It's best to double check the global results with beta-diversity metrics such as permanova.

Visualizing the raw differentials is fine. The intercept just represents the average population for your control group (which I'm assuming you aren't interested in). If you haven't already, I recommend checking out this tool to help visualizing differentials.

The results I get using Songbird are similar with results I get when e.g. comparing the taxa barplots. However, most of the times my null model reports a Q2 Score of below zero (e.g. -0.002). Again, I generally dont see any major differences between my treatments using beta diversity tests either. Is it still possible to use the songbird differentials or not at all?