Help with ANCOM result interpretation - "Reject null hypothesis" is true for all features


I’m having difficulties understanding ANCOM output (using Qiime2-2017.10). The ancom.csv shows true for all features in the “reject null hypothesis” column while the W-stats show 0 for most of the features. Does it mean that all of the features are differentially abundant (highly unlikely) even though W = 0?

Here is a snapshot from ancom.csv :

	Reject null hypothesis	W
3d80710fb18afb303baf125753fd7ac8	True	0
d55aec9ec60a4e3097c3c3953dae8773	True	1
f80a17436037d706dc728dd265b7c21e	True	2
f5487404313a4fb57bc3489c167e9a37	True	0
b3fef1e26b7b6ac20847ede4da68547a	True	0
ba53aaf1e742e7d9bf5d5f336ca52ca9	True	0
d3211b7a04e5fb36a383450b8f5710e2	True	0
08c92baa0290e451ea255f6020e44739	True	1
6bc890a2012bca332b9d5c43905d1aae	True	0
3af2197bbd2daf9be46127b58559c91f	True	3
41b518f17927240fdfd1465cd351afc4	True	0
10df701d47286a6ed30c544e219c87f4	True	1
2eb7bab7b0af702e4f18358968a8f7ff	True	0
7a08a37eff6224b432a537907a0cec2c	True	0
27e5a0119bf0f99608903f79f07ce884	True	0
310e1b0125888ac49bf01d4fc73500d5	True	0
1141f30eddb626a3a1151a19445eadfe	True	0
c2cac7ce4200877d49ef4a0f53183000	True	0

Thanks in advance!

That is an indication that the threshold is not behaving as expected. If W is 0, then that should not be detected to be significantly different. I would aim to only investigate the species with high W scores (and definitely not those that are zero).

Btw, I gave an explanation of how the W statistic works here. Hopefully that clarifies how this statistic works!

1 Like

Alright. The explanation of how W-stats work is helpful. One other question, is there a guide for the cut-off of W-stats for significance?

The cut-off for the W-stats is automated, there are a few parameters that can be tuned within scikit-bio. But there isn’t a guide that I know of to tune these parameters.

It maybe worthwhile to contact Shyamal Peddada - the original author of this method.

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