# ANCOM, testing assumptions, filtering

**URL:** https://forum.qiime2.org/t/ancom-testing-assumptions-filtering/2752
**Category:** User Support
**Tags:** ancom
**Created:** [January 26, 2018, 11:30pm UTC](https://forum.qiime2.org/t/ancom-testing-assumptions-filtering/2752 "2018-01-26T23:30:03Z")
**Posts on this page:** 1
**Showing post:** 3

<div class="post-metadata">

### Author: ![Mehrbod\_Estaki](https://forum.qiime2.org/user_avatar/forum.qiime2.org/mehrbod_estaki/32/4001_2.png) [@Mehrbod\_Estaki](https://forum.qiime2.org/u/Mehrbod_Estaki)
#### Post date: [January 29, 2018, 8:55pm UTC](https://forum.qiime2.org/t/ancom-testing-assumptions-filtering/2752/3 "2018-01-29T20:55:56Z")

</div>

Great question! I'm also curious as to how to test the [assumptions](https://docs.qiime2.org/2017.12/tutorials/moving-pictures/#differential-abundance-testing-with-ancom) for ANCOM (i.e less than 25% of features changing between groups). An extension to the question, how does this exactly work across multiple groups? 25% change across pairs or overall?

@jessicalmetcalf, from this [thread](https://forum.qiime2.org/t/ancom-giving-strange-w-values/1002/3) it sounds like too many zeros or low counts are better off removed:

> Low counts / zeros could cause false positives in ANCOM

---

_[View the full topic](https://forum.qiime2.org/t/ancom-testing-assumptions-filtering/2752)._