I wanted to find a new method to reanalyze my 16S amplicon data. So I searched in this forum, and I find songbird. I read the paper of songbird and thought it would be a good technique for me. But as we know, songbird provides differential ranks and no p value or adjusted p value and I just don’t know how to choose significant differential taxa. That means I have to do null hypothesis test by myself. So my question is how to conduct the null hypothesis test, using which statistical method. And what is the proper criteria to find significant differential taxa?
My another question is whether the coefficient provided by songbird is log(a/b)? a is the ratio of abudance of genus x in a condition and reference microbe; b is the ratio of abundance of genus x in b condition and reference microbe.
Thank you very much!
Hi @Zhanzhan, you are right. Null hypothesis testing is really hard.
I’d check out qurro - that can help you explore your differentials.
@mortonjt Thank you very much! I go through the paper of qurro and know that qurro is a visualization tool. That is to say I can use qurro to visualize the results of songbird. And I want to know if I can combine the results of songbird and ALDEx2 in a report. ALDEx2 is performed to identify significant differential taxa and songbird is used to identify the most increasing and decreasing taxa in two conditions. I can’t determine if it is ok. Could you please give me some advise?
No need to combine aldex2 and songbird. If you only have categorical variables, you can pass in the results of aldex2 directly into qurro (aldex2 and songbird both implement multinomial regression).
One thing to note about aldex2 is that it chooses the mean taxa to be the reference frame. If your sample mean isn’t changing, it is fine - but if it is drastically changing (i.e. antibiotics usage, tooth brushing, …), then this will drastically skew your pvalues.
This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.