# Mixture model in Pumas

**URL:** <https://discourse.pumas.ai/t/mixture-model-in-pumas/347>\
**Category:** How-to\
**Created:** [May 17, 2020, 6:37am UTC](https://discourse.pumas.ai/t/mixture-model-in-pumas/347 "2020-05-17T06:37:18Z")\
**Posts on this page:** 8\
**Page:** 1

<div class="post-metadata">

**Author:** ![ashwin](https://avatars.discourse-cdn.com/v4/letter/a/f08c70/32.png) [@ashwin](https://discourse.pumas.ai/u/ashwin)\
**Post date:** [May 17, 2020, 6:37am UTC](https://discourse.pumas.ai/t/mixture-model-in-pumas/347/1 "2020-05-17T06:37:18Z")

</div>

Hi all, does Pumas have mixture model functionality? I’m haven’t found anything in the documentation with regards to it. Thanks!

---

<div class="post-metadata">

**Author:** ![vijay](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.pumas.ai/vijay/32/96_2.png) [@vijay](https://discourse.pumas.ai/u/vijay)\
**Post date:** [May 18, 2020, 8:39am UTC](https://discourse.pumas.ai/t/mixture-model-in-pumas/347/2 "2020-05-18T08:39:30Z")

</div>

Welcome to the community @ashwin. The current release does not include Mixture Model handling, but this is planned and will be in an upcoming release.

---

<div class="post-metadata">

**Author:** ![MasatoFukae](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.pumas.ai/masatofukae/32/341_2.png) [@MasatoFukae](https://discourse.pumas.ai/u/MasatoFukae)\
**Post date:** [June 13, 2023, 9:16pm UTC](https://discourse.pumas.ai/t/mixture-model-in-pumas/347/3 "2023-06-13T21:16:54Z")

</div>

Hi all. I also need mixture model functionality. Is it available now?

---

<div class="post-metadata">

**Author:** ![pkofod](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.pumas.ai/pkofod/32/59_2.png) [@pkofod](https://discourse.pumas.ai/u/pkofod)\
**Post date:** [June 14, 2023, 7:31am UTC](https://discourse.pumas.ai/t/mixture-model-in-pumas/347/4 "2023-06-14T07:31:58Z")

</div>

Dear Masato Fukae,

Unfortunately, it is not yet available.

Best,  
Patrick

---

<div class="post-metadata">

**Author:** ![MasatoFukae](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.pumas.ai/masatofukae/32/341_2.png) [@MasatoFukae](https://discourse.pumas.ai/u/MasatoFukae)\
**Post date:** [June 22, 2023, 12:59am UTC](https://discourse.pumas.ai/t/mixture-model-in-pumas/347/5 "2023-06-22T00:59:24Z")

</div>

Hi Pumas community,

I am still trying to make a mixture of sub-populations somehow. I expected `Categorical` in the `@random` block to work, but it did not. Is there some workaround? It would be great if anyone gives me an idea.

Thanks,  
Masato

---

<div class="post-metadata">

**Author:** ![MasatoFukae](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.pumas.ai/masatofukae/32/341_2.png) [@MasatoFukae](https://discourse.pumas.ai/u/MasatoFukae)\
**Post date:** [June 22, 2023, 1:02am UTC](https://discourse.pumas.ai/t/mixture-model-in-pumas/347/6 "2023-06-22T01:02:43Z")

</div>

Let me post a code fragment to be more specific.

```auto
    @random begin
        ηkd ~ Normal(0, sqrt(Ωkd))
        ηbase ~ MvNormal(Ωbase)
        # subpop ~ Categorical(p1, p2, p3, p4)
        subpop ~ Categorical(p1, p2)
    end

```

---

<div class="post-metadata">

**Author:** ![storopoli](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.pumas.ai/storopoli/32/190_2.png) [@storopoli](https://discourse.pumas.ai/u/storopoli)\
**Post date:** [June 22, 2023, 1:23pm UTC](https://discourse.pumas.ai/t/mixture-model-in-pumas/347/7 "2023-06-22T13:23:29Z")

</div>

> [@MasatoFukae](#):
>
> I expected `Categorical` in the `@random` block to work, but it did not.

I don’t think `Categorical` is supported. The list of supported distributions are at [Defining NLME models in Pumas · Pumas](https://docs.pumas.ai/stable/basics/models/).

---

<div class="post-metadata">

**Author:** ![MasatoFukae](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.pumas.ai/masatofukae/32/341_2.png) [@MasatoFukae](https://discourse.pumas.ai/u/MasatoFukae)\
**Post date:** [June 22, 2023, 4:16pm UTC](https://discourse.pumas.ai/t/mixture-model-in-pumas/347/8 "2023-06-22T16:16:15Z")

</div>

Thank you for your quick response. I was trying it because I found it here [PumasModel-Error models · Pumas](https://docs.pumas.ai/stable/model_components/error_models/#Summary-of-Error-Models). But maybe available distributions are different between `@random` and `@derived`. And only continuous ones are ready in `@random` so far. Is my understanding correct? If so, do you have some ideas about how to generate discrete- or bi-modal-like distributions using currently available ones?