F_onewayConstantInputWarning Error in ANCOM

Technical Support
#1

Hi,

I just had a quick question about ANCOM analysis through QIIME. I’ve been trying to run ANCOM on my taxonomic data (to see whether there are significant differences between 3 categories of patients: healthy, treated, and untreated/diseased). I used this command:

qiime composition ancom \

–i-table comp-table.qza
–m-metadata-file metadata.tsv
–m-metadata-column subject-type
–o-visualization ancom-subject-type.qzv
–verbose

Although I did get an output, I also got the F_onewayConstantInputWarning error:
/Users/lyskae/opt/anaconda3/envs/qiime2-2020.11/lib/python3.6/site-packages/scipy/stats/stats.py:3641: F_onewayConstantInputWarning: Each of the input arrays is constant;the F statistic is not defined or infinite
warnings.warn(F_onewayConstantInputWarning())

I’m unsure as to how to fix this, and whether this affects the analysis. Any help would be greatly appreciated!

#3

Welcome to the forum, @lys.kae, and neat warning! I’ve never seen this one before.

It seems to be bubbling up from scipy’s one-way ANOVA implementation, and is warning you that your data may be degenerate (in ways that could affect the meaningfulness of your results).

Specifically, if each of the “groups” that is being compared contains identical data values, the ANOVA test used by ANCOM will generate “garbage” data. (That is, apparently, the technical term)

If I understand the issue correctly, you may have two samples in your feature table with identical feature counts. Is there any chance that one or more samples in your feature table have been accidentally duplicated (under different sample ids)? This could also happen by chance (likelier in very-low-biomass samples), but it seems improbable.

Let us know what you find!
Chris :boar: