I have three groups of individuals and a sample of them had their microbiota sequenced at 2 days , and a different sample at 6d. However, the samples are not paired i.e. the same individual was not sampled twice.
I would like to run analysis to determine whether shannon diversity changed significantly over time according to the group. Is there a way to do this on unpaired data? I’m interested in the q2 longitudinal but that seems only for paired data
Is this what your sample metadata looks like?
| sampleID | group | day | Person |
| s1 | groupA | 2 | person1 |
| s2 | groupA | 6 | person2 |
| s3 | groupB | 2 | person3 |
| s4 | groupB | 6 | person4 |
| s5 | groupC | 2 | person5 |
| s6 | groupC | 6 | person6 |
Because the people are not paired, you could do a simple blocked analysis comparing group (A vs B vs C) or day (2 vs 6), or the interactive effect between both group*day.
One way to do that is using alpha-group-significance as shown here, but I would check in with a statistician to see how they would handle this study. How did you plan to control for person-to-person variation? Were the people randomly assigned to these three groups?
Thanks for your reply! Yes, my metadata does look something like that. Individuals were first assigned to groups based on a phenotype and then randomly selected for sequencing. I’ve already done the alpha group significance but was wondering if there was anything else I could do.
Are these… chickens? Am I mixing you up with someone else?
Mosquitoes! Hence why I can’t sample the same individual twice would be amazing if I could though
Oh it’s an animal model that does not allow paired sampling! I should have asked about that.
Because your subjects were randomly selected for timepoint and sequencing, I think your use of alpha group significance makes good sense. This study requires destructive sampling, and your study design takes care of that!
When I worked on this paper, I dealt with this same issue; samples could only be taken once, yielding unpaired data. We started by looking at alpha diversity over time (Fig. 1), then added an analysis of nestedness (species loss) and turnover (species replacement) that made sense for our study (Fig. 3).
Even when paired analysis were not on the table, we found a good way to answer the underlying biological question.
What are you investigating in your study?
Sorry, I think I had you confused with someone else who had coprophagic chickens. . So, then, it sounds like you can’t trick the system using something like that. Then, I think @colinbrislawn’s excellent suggestion is the way to go.
This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.