# stepwise text

**URL:** <https://forum.certara.com/t/stepwise-text/3622>\
**Category:** Modelling and Simulation\
**Created:** [July 5, 2022, 1:18am UTC](https://forum.certara.com/t/stepwise-text/3622 "2022-07-05T01:18:22Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![crystal](https://avatars.discourse-cdn.com/v4/letter/c/2bfe46/32.png) [@crystal](https://forum.certara.com/u/crystal)\
**Post date:** [July 5, 2022, 1:18am UTC](https://forum.certara.com/t/stepwise-text/3622/1 "2022-07-05T01:18:22Z")

</div>

Dear all,

I have a problem about the Stepwise run model. I built a POP PK model, after I finished covariates stepwise, everything seems to be good, but when I check the Results—TextOutput—Stepwise Text, it confuses me that the criterion of stepwise forward turn from 3.841 into 5.991 from cstep010 to cstep015（Add P value=0.05）? Does anyone know why?

Thanks a lot.

---

<div class="post-metadata">

**Author:** ![cradhakr](https://avatars.discourse-cdn.com/v4/letter/c/7feea3/32.png) [@cradhakr](https://forum.certara.com/u/cradhakr)\
**Post date:** [July 5, 2022, 3:54am UTC](https://forum.certara.com/t/stepwise-text/3622/2 "2022-07-05T03:54:42Z")

</div>

Hi Crystal,

This is the explanation for your question

• The LRT compares two likelihoods • For nested models the difference of two log likelihoods is asymptotically chi-squared distributed: – For 1 parameter difference (1 degree of freedom) the critical value from the chi-squared distribution is ‘3.84’ (P=0.05) \

[http://holford.fmhs.auckland.ac.nz/docs/principles-of-covariate-modelling.pdf](http://holford.fmhs.auckland.ac.nz/docs/principles-of-covariate-modelling.pdf)

The attached link could be of use to you.

Thanks

Chandramouli (Mouli)

---

<div class="post-metadata">

**Author:** ![crystal](https://avatars.discourse-cdn.com/v4/letter/c/2bfe46/32.png) [@crystal](https://forum.certara.com/u/crystal)\
**Post date:** [July 5, 2022, 6:19am UTC](https://forum.certara.com/t/stepwise-text/3622/3 "2022-07-05T06:19:25Z")

</div>

> [@cradhakr](#):
>
> Hi Crystal,
> 
> This is the explanation for your question
> 
> • The LRT compares two likelihoods • For nested models the difference of two log likelihoods is asymptotically chi-squared distributed: – For 1 parameter difference (1 degree of freedom) the critical value from the chi-squared distribution is ‘3.84’ (P=0.05) \
> 
> [http://holford.fmhs.auckland.ac.nz/docs/principles-of-covariate-modelling.pdf](http://holford.fmhs.auckland.ac.nz/docs/principles-of-covariate-modelling.pdf)
> 
> The attached link could be of use to you.
> 
> Thanks
> 
> Chandramouli (Mouli)

Thanks, Chandramouli for your help, I really appreciate it.

---

<div class="post-metadata">

**Author:** ![cradhakr](https://avatars.discourse-cdn.com/v4/letter/c/7feea3/32.png) [@cradhakr](https://forum.certara.com/u/cradhakr)\
**Post date:** [July 5, 2022, 8:36am UTC](https://forum.certara.com/t/stepwise-text/3622/4 "2022-07-05T08:36:15Z")

</div>

HI Crystal,

Forgot to add the explanation - With RACE you might have 2 levels and the p value is 0.05

with

df =2 it is - 5.991465.

df=1 it is - 3.841459

Thanks

Mouli

---

<div class="post-metadata">

**Author:** ![Samer\_Mouksassi](https://avatars.discourse-cdn.com/v4/letter/s/ce7236/32.png) [@Samer\_Mouksassi](https://forum.certara.com/u/Samer_Mouksassi)\
**Post date:** [July 5, 2022, 8:39am UTC](https://forum.certara.com/t/stepwise-text/3622/5 "2022-07-05T08:39:43Z")

</div>

you can use R to know what are the thresholds:

> qchisq(1-0.05,1:5)  
> [1] 3.841459 5.991465 7.814728 9.487729 11.070498  
> qchisq(1-0.01,1:5)  
> [1] 6.634897 9.210340 11.344867 13.276704 15.086272  
> qchisq(1-0.001,1:5)  
> [1] 10.82757 13.81551 16.26624 18.46683 20.51501

the above cover p = 0.05, 0.01 and 0.001  
for 1 to 5 degrees of freedom

Samer