# Reporting shrinkage for IOV

**URL:** <https://forum.certara.com/t/reporting-shrinkage-for-iov/3709>\
**Category:** Modelling and Simulation\
**Created:** [October 13, 2023, 4:42am UTC](https://forum.certara.com/t/reporting-shrinkage-for-iov/3709 "2023-10-13T04:42:23Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![cxzhang](https://avatars.discourse-cdn.com/v4/letter/c/6f9a4e/32.png) [@cxzhang](https://forum.certara.com/u/cxzhang)\
**Post date:** [October 13, 2023, 4:42am UTC](https://forum.certara.com/t/reporting-shrinkage-for-iov/3709/1 "2023-10-13T04:42:23Z")

</div>

Dear all,

For a popPK model I am working on, I estimated interoccasion variability (IOV) for CL for 4 different occasions, assuming the IOV estimates to be the same for all 4 occasions due to limited data availability. Therefore, I obtained 4 identical omega estimates and 4 identical standard error estimates for the four occasions (i.e. IOV on CL for occasion 1 = IOV on CL for occasion 2 = IOV on CL for occasion 3 = IOV on CL for occasion 4). My understanding is that I only need to report one estimate and its standard error for IOV on CL as my model output, instead of reporting the same set of values 4 times defined by the occasions. However, I do not know what to report for the shrinkage associated with IOV on CL. In the Omega sheet, 4 different shrinkage values were shown associated with the 4 occasions, ranging from 0.3 to 0.9. From looking at published studies, people report only one shrinkage value for each IOV tested. How can I obtain an overall shrinkage value to report for the IOV on CL? The number of data points for each occasion are different (e.g. intense sampling on occasion 1 and 4, sparse sampling on occasion 2 and 3), in case this matters.

Thank you for your help!

---

<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:** [October 13, 2023, 9:10am UTC](https://forum.certara.com/t/reporting-shrinkage-for-iov/3709/2 "2023-10-13T09:10:55Z")

</div>

while IOV OCC 1, 2 , 3.. have a common variance

the posthocs will be different i.e the individual occasion specific etaclocc1 , will be different and hence you will get

and occasion specific shrinkage

not sure it makes sense to compute an overall shrinkage over all occasions

---

<div class="post-metadata">

**Author:** ![cxzhang](https://avatars.discourse-cdn.com/v4/letter/c/6f9a4e/32.png) [@cxzhang](https://forum.certara.com/u/cxzhang)\
**Post date:** [October 13, 2023, 5:12pm UTC](https://forum.certara.com/t/reporting-shrinkage-for-iov/3709/3 "2023-10-13T17:12:56Z")

</div>

> [@smouksassi1](#):
>
> while IOV OCC 1, 2 , 3.. have a common variance
> 
> the posthocs will be different i.e the individual occasion specific etaclocc1 , will be different and hence you will get
> 
> and occasion specific shrinkage
> 
> not sure it makes sense to compute an overall shrinkage over all occasions

Thank you smoukasassi1 for your reply! Yes, I do agree with you. I understand why there are separate shrinkage values for each occasion. Therefore, I am very confused by what I’ve seen in popPK literature because when IOV shrinkage is reported, only 1 value is usually presented, such as in the example below (Chan et al., Pharm Res. 2020 Jan 6;37(2):25.

doi: 10.1007/s11095-019-2752-y.).

For IOV to make sense, there have to be multiple occasions defined. Therefore, several shrinkage values should have been estimated for each occasion. I should probably rephrase my question:

1. Why do people report only one shrinkage value for each IOV? What is this single shrinkage value telling us and is this standard practice?

2. Is there a way to obtain a single shrinkage value?

3. Or is this due to some algorithm differences in how IOV is calculated between NONMEM and Phoenix NLME?

Thank you!

 ![Screenshot 2023-10-13 at 9.58.56 AM.png](https://us1.discourse-cdn.com/certara/original/2X/7/788e59b60278b6121f182bb19f306f90e8e78d5f.png)