Different p-values with PERMANOVA after editing metadata

Hi all,

I have a question regarding my results. I ran the beta diversity group significance command to compare groups with a different metadata category. I added one column to the metadata file (not the one presented here) and rerun the analysis from start. However, now after running the commands again I got different p-values. Any idea what would be the issue? The old results are analyzed with Qiime2 version 2019-4.

new results.qzv (386.8 KB)
old result.qzv (387.1 KB)

Hi @hennihi,

So I understand your pipeline, what step did you re-run from when you changed your metadata? (The feature table? Diversity? Testing?)

If you re-did from diversity-core metrics, then you re-rarified the table (which is a stochastic process) and then re-ran a permutative test (which is also a random process). So, you have two random processes that resulted in a slight difference in your results, but on the whole, give very similar numbers (167/1000 permutations were more extreme before and 188/1000 permutations were more extreme in the new results).

However, you don’t need to re-run your data when you edit your metadata. (God knows, if you did a lot of us would be in trouble!) There was a long discussion on the topic a while ago; it might be worth checking out to get a better basis for what’s going on under the hood and to help figure out which steps have to be re-done where.

Mistake in metadata file and re-running core metrics analysis.

Best,
Justine

1 Like

Thank you for a fast reply and assuring me this is okay! I re-did the core-diversity metrics. Thank you for your opinion and the link you provided, it made me understand the process better. I’m sure it will be useful in the future.

1 Like