# The P values generated by Maaslin3

**URL:** <https://forum.biobakery.org/t/the-p-values-generated-by-maaslin3/9010>\
**Category:** MaAsLin\
**Created:** [July 27, 2026, 7:27am UTC](https://forum.biobakery.org/t/the-p-values-generated-by-maaslin3/9010 "2026-07-27T07:27:07Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![yuxuanzhao](https://avatars.discourse-cdn.com/v4/letter/y/45deac/32.png) [@yuxuanzhao](https://forum.biobakery.org/u/yuxuanzhao)\
**Post date:** [July 27, 2026, 7:27am UTC](https://forum.biobakery.org/t/the-p-values-generated-by-maaslin3/9010/1 "2026-07-27T07:27:07Z")

</div>

Hi,

Firstly, thank you so much for developing Maaslin3! We have done some analyses on species abundance using Maaslin3 and we had a few questions regarding the significant outputs:

1. If we want to draw forest plot using the significant outputs, is it correct to generate the 95% confidence intervals manually by “coef ± 1.96\*stderr”?

2. We have noticed that in some of our outputs, the P values and Q values denote significant results but the CIs contain 0 when we tried to generate CIs (as shown in the picture below). We are wondering whether the P values have been adjusted by other methods or we are using the wrong formula to generate CIs.

Thanks a lot for your help!

---

<div class="post-metadata">

**Author:** ![WillNickols](https://yyz1.discourse-cdn.com/flex027/user_avatar/forum.biobakery.org/willnickols/32/3223_2.png) [@WillNickols](https://forum.biobakery.org/u/WillNickols)\
**Post date:** [July 27, 2026, 9:18pm UTC](https://forum.biobakery.org/t/the-p-values-generated-by-maaslin3/9010/2 "2026-07-27T21:18:27Z")

</div>

Hi,

1. Yes, that would produce the correct confidence intervals for the relative abundance associations. (Note that these won’t be FDR-corrected, but most people just don’t FDR correct confidence intervals.)
2. By default, MaAsLin compares coefficients to the median of that metadatum’s coefficients in the hypothesis test in order to test for absolute abundance associations. (The paper has more details on why and how this works.) This means that instead of the usual hypothesis test against 0, it’s testing against the value in the `null_hypothesis` column. If you don’t care about the absolute abundance associations and just want the relative abundance associations, you can set `median_comparison_abundance` to `FALSE` and it’ll hypothesis test against 0. This will give you the expected results with CIs containing 0 iff the association is non-significant.

Will

---

<div class="post-metadata">

**Author:** ![yuxuanzhao](https://avatars.discourse-cdn.com/v4/letter/y/45deac/32.png) [@yuxuanzhao](https://forum.biobakery.org/u/yuxuanzhao)\
**Post date:** [July 28, 2026, 1:16am UTC](https://forum.biobakery.org/t/the-p-values-generated-by-maaslin3/9010/3 "2026-07-28T01:16:05Z")

</div>

Thank you so much for the explanation!