I have just found that the different results were not given by the command for shannon index, but by "qiime diversity core-metrics-phylogenetic". I ran 3 times with this command, each time i produced shannon_group.qzv, i got different p-values. Seems the final conclusion is not changed (at least for my experiment), but i am wondering why is that.
I upload 3 shannon indexes from 3 different runs here. please choose "Timepoint" category to check.
Ah, well then you have your answer! This command will rarefy your table as part of the processing, before computing the alpha and beta diversity metrics. Each time you rarefy you are potentially going to see a different result --- this is why commands like diversity alpha-rarefaction and diversity beta-rarefaction exist, to help you understand the impact of even sampling on these downstream computations.
Sorry, I got confused again. I ran the same command “qiime diversity core-metrics-phylogenetic” 3 times with the same “sampling depth”, It means that i have thrown the same number of sequences. why the result can be different?
Same # of sequences, but different sequences. Rarefy == randomly sampling N sequences from among a larger population of sequences. So each time you rarefy you get a different subsample, leading to distinct results.
Thanks a lot for your reply! yes, it makes sense. But Should i run the commands several times to make sure that the conclusions are correct. And how many times?
No, you do not need to. If results differ, it is probably because you are rarefying at a low depth. See @thermokarst's advice above about determining how rarefaction depth impacts results:
You can choose an appropriate rarefaction depth based on those results.