How to report ANCOM results for manuscript

[mweinroth (Maggie Weinroth)]

I have run ANCOM on some of my data and am running into trouble on what the correct way to report the results is. From the attached qzv, would this be an appropriate sentence:

The phyla Verrucomicrobia differed by region (f-statistic = 16.5, W = 10) as the high plains region had a median of of 66.5 sequences, the south had 381.5, and the west had 885.5.

l2_ancom_region_soil_6418.qzv (314.6 KB)

[mweinroth (Maggie Weinroth)]

@mortonjt just bumping this to see if you had any thoughts.

Thanks!

[mortonjt (Jamie Morton)]

I wouldn’t include the f-statistic, that is just used for visualization. W = 10 is fine. May be worthwhile to note that when W=10, the phyla Verrucomicrobia was detected to be significantly different relative to 10 other phyla.

[mweinroth (Maggie Weinroth)]

I was under the impress that this result was telling me there was a difference in abundance of Verrucomicrobia between my treatments–is this accurate?

So I am not sure what you mean by different from 10 other phyla, can you explain that further? Is there a way to know what 10 phyla it is significantly different from? Also, does then mean that it is not significantly different than the other 22?

[mortonjt (Jamie Morton)]

The f-statistic is just for visualization purposes to look at differences in relative abundances. Differences in relative abundances do not necessary reflect differences in absolute abundances. So this is not a statistic that I would recommend reporting.

Here are the posts where the ANCOM W-statistic is discussed in more detail

Ancom significance

Hi @nricks, @Mehrbod_Estaki raises many great points (thank you!), including how the W values are calculated. Thus W will depend on the number of features, and you cannot really compare W values between ANCOM tests on two different feature tables. Different threshold W scores will be set for each test. So what you are seeing is not really a “problem”. However, what is a problem (as @Mehrbod_Estaki points out) is that in table #2 many many features are significantly different, which is breaki…

Specify W cutoff for anacom?

group_significance.py link is fixed!

Not necessarily - just because you have 10 phyla that are significantly different, it doesn't mean that the other 22 phyla aren't significant. It just means that you couldn't rule out the possibly that they aren't significant. Maybe worthwhile reading up more about statistical hypothesis testing in addition to the supplemental section in the ANCOM paper

[aeriel.belk]

I have a similar question on this. If my research question is whether a phyla changes in relative abundance when a treatment is applied, it seems like the W value is not helpful. However, the visualization does not provide a p-value, which would be most useful for a hypothesis test and, I believe, work better when reporting the statistics in a manuscript. Is there a way to get a p-value when running this test?

[mortonjt (Jamie Morton)]

ANCOM does provide a threshold for accepting / rejecting if a taxa abundance has been altered. This should be good enough for reporting.

But in the grand scheme of things, it is never possible to provide a useful p-value in the first place since it is not possible to construct a null hypothesis. I explain some of the intuition behind this in my paper here.