Using qiime longitudinal pairwise-differences, I can obtain the results for pairwise difference tests (evaluate changes in a specific metric using paired samples) and Pairwise group comparison tests (compare the delta of these changes according to a metadata column).
Using qiime longitudinal pairwise-distances, I can only see Pairwise group comparison tests (group A vs. group B). How can I know whether the distance matrix changed over time among samples from the same group? Maybe Iām missing something.
You are not missing anything. The wilcoxon signed-rank test for difference from 0 is only used for pairwise differences.
One reason it is not implemented in pairwise-distances is that in real samples the distance between two samples will almost never be 0, and cannot be negative, so such a test would almost always be significant but the results are meaningless.
This test makes much more sense with pairwise-differences, as alpha diversity data and other inputs to this action can change in both directions (or are likely to be ~0 in real datasets), so a significant test results is much more meaningful, indicating a directional change.
Thanks for all your support. I got significant results using pairwise distance comparisons function between two groups in my data. Is there any way I can visualize which taxa were responsible for these results?
If this yields a short list of taxa, you can actually use pairwise-differences (with a FeatureTable[RelativeFrequency] as input) to plot boxplots of taxonomic abundance for specific taxa (one at a time). Just ignore the stats ā use ANCOM for significance (the tests used by that method are not suitable for relative frequency data).
If ANCOM yields a long list of significant taxa and you are interested in plotting all together, you will need to filter your feature table and use as input to qiime taxa barplot or you may just need to export and do some custom plotting, e.g., with R.