# P-values- backward model selection gneiss

**URL:** https://forum.qiime2.org/t/p-values-backward-model-selection-gneiss/5395
**Category:** User Support
**Tags:** gneiss
**Created:** [August 3, 2018, 4:55am UTC](https://forum.qiime2.org/t/p-values-backward-model-selection-gneiss/5395 "2018-08-03T04:55:12Z")
**Posts on this page:** 6
**Page:** 1

<div class="post-metadata">

### Author: ![Fabs](https://forum.qiime2.org/letter_avatar_proxy/v4/letter/f/74df32/32.png) [@Fabs](https://forum.qiime2.org/u/Fabs)
#### Post date: [August 3, 2018, 4:55am UTC](https://forum.qiime2.org/t/p-values-backward-model-selection-gneiss/5395/1 "2018-08-03T04:55:12Z")

</div>

Hello,

I have a quick question. If I am running a backward model selection and I want to use the P-values provided in Gneiss output to see which variables to remove to create my model. The problem is, I've downloaded the P-values but there are quite a few yo, y1, y2..... so I am unsure which and how to use this table for model selection.

Would you mind explaining the table and letting me know how to use it.

Please see the results of the model, containing all variables.

 ![image](https://forum-qiime2-org.s3.dualstack.us-west-2.amazonaws.com/original/2X/c/cb49c592f93dbd171d20500c60f60c6e48a4157f.png)

P-VALUES

 ![image](https://forum-qiime2-org.s3.dualstack.us-west-2.amazonaws.com/original/2X/3/316b37a5198e4936967a22c714202acc8d0a67af.png)

Thanks 🙂

---

<div class="post-metadata">

### Author: ![Nicholas\_Bokulich](https://forum.qiime2.org/user_avatar/forum.qiime2.org/nicholas_bokulich/32/19937_2.png) [@Nicholas\_Bokulich](https://forum.qiime2.org/u/Nicholas_Bokulich)
#### Post date: [August 3, 2018, 1:29pm UTC](https://forum.qiime2.org/t/p-values-backward-model-selection-gneiss/5395/2 "2018-08-03T13:29:39Z")

</div>



---

<div class="post-metadata">

### Author: ![mortonjt](https://forum.qiime2.org/user_avatar/forum.qiime2.org/mortonjt/32/3488_2.png) [@mortonjt](https://forum.qiime2.org/u/mortonjt)
#### Post date: [August 3, 2018, 5:21pm UTC](https://forum.qiime2.org/t/p-values-backward-model-selection-gneiss/5395/3 "2018-08-03T17:21:00Z")

</div>

Hi @Fabs,

I wouldn't use the p-values for model selection, the MSE or the Rsquared calculations is a much better statistic for this, since it is summarizing over the entire model.

---

<div class="post-metadata">

### Author: ![Nicholas\_Bokulich](https://forum.qiime2.org/user_avatar/forum.qiime2.org/nicholas_bokulich/32/19937_2.png) [@Nicholas\_Bokulich](https://forum.qiime2.org/u/Nicholas_Bokulich)
#### Post date: [August 3, 2018, 6:04pm UTC](https://forum.qiime2.org/t/p-values-backward-model-selection-gneiss/5395/4 "2018-08-03T18:04:48Z")

</div>



---

<div class="post-metadata">

### Author: ![Fabs](https://forum.qiime2.org/letter_avatar_proxy/v4/letter/f/74df32/32.png) [@Fabs](https://forum.qiime2.org/u/Fabs)
#### Post date: [August 4, 2018, 5:36am UTC](https://forum.qiime2.org/t/p-values-backward-model-selection-gneiss/5395/5 "2018-08-04T05:36:45Z")

</div>

Perfect! Thanks for letting me know.

---

<div class="post-metadata">

### Author: ![system](https://forum-qiime2-org.s3.dualstack.us-west-2.amazonaws.com/original/3X/2/1/21af5fe23cb6f4579467c66a9ed94e55274ca7bd.svg) [@system](https://forum.qiime2.org/u/system)
#### Post date: [September 4, 2018, 11:36am UTC](https://forum.qiime2.org/t/p-values-backward-model-selection-gneiss/5395/6 "2018-09-04T11:36:51Z")

</div>

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