The difference between beta-group-significance q2-longitudinal and beta-group-significance

Could you please explain the difference between the analysis and beta-group-significance. For example, I sampled from the same reactor at day0, day 7and day14. I do the beta-group-significance analysis for the matrix.qza, and should I must do the analysis “q2-longitudinal”? I can not understand their difference, because these two analysis can both do the pair-wise comparsion.

Good morning,

Great question! While these plugins both test differences in beta diversity, they make a different assumption about the groups they test.

The beta-group-significance plugin assumes independent categorical data.
The q2-longitudinal plugin assumes continuous data (and maybe directional data).

So because your samples have a order (0,7,14 is not the same as something random like 14,0,7), you should use the q2-longitudinal plugin.

Does that help answer your question? :qiime2:



Thank you for your reply.

If I want to know which factor has the largest source of variation in the factors tested, should I still use qiime diversity adonis?

Please check the attached metadata. If I want to test whether distances between samples of "high pressure", are more similar to each other than they are to samples of "low pressure", could you I use qiime diversity beta-group-significance and --m-metadata-column Pressure? Because I think the "Pressure" is not in chronological order, while the "Batch" is in chronological order.

metadata_withoutbg.txt (3.2 KB)

1 Like

Thanks for telling me more!

Sounds like a good plan. I also use the adonis test when I’m interested in questions like this.

So the adonis test does not care at all about order, when dealing with a factor. Text strings like High and Low are interpreted as factors, while numbers are dealt with as continuous variables. Based on the metadata you provided, I think the adonis test would compare your Date variable as a factor, which is a reasonable thing to do and would let you compare 1_2 vs 1_14, but it might be more appropriate to treat date as a single continuous variable (1 -> 14 -> 26).

In some ways this comes down to the stats question of treating time as a discreet factor or continuous number. And that depends on how you want to frame your scientific question. As with a lot of stats, is no obvious right answer to this question and you get to choose how you frame this comparison. 🤷


Thank you for your clear answers! :100:

1 Like

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.