# Can we access the component distributions in a mixture model?

**URL:** <https://discourse.pymc.io/t/can-we-access-the-component-distributions-in-a-mixture-model/3361>\
**Category:** Questions\
**Created:** [May 31, 2019, 3:34am UTC](https://discourse.pymc.io/t/can-we-access-the-component-distributions-in-a-mixture-model/3361 "2019-05-31T03:34:28Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![rpgoldman](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/rpgoldman/32/1800_2.png) [@rpgoldman](https://discourse.pymc.io/u/rpgoldman)\
**Post date:** [May 31, 2019, 3:34am UTC](https://discourse.pymc.io/t/can-we-access-the-component-distributions-in-a-mixture-model/3361/1 "2019-05-31T03:34:28Z")

</div>

If I have a mixture model with a Gaussian (for the central tendency) and a Uniform (for the outliers), is there some way I can look at the samples of the component distributions? Since they are distributions, they don’t have names…  
If I make the mixture model “by hand” (as in @junpenglao’s schizophrenic reaction time example), I can “look into” the components of a mixture model, but the built in `pm.MixtureModel` doesn’t seem to support this in any way.

---

<div class="post-metadata">

**Author:** ![junpenglao](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/junpenglao/32/8_2.png) [@junpenglao](https://discourse.pymc.io/u/junpenglao)\
**Post date:** [May 31, 2019, 5:14am UTC](https://discourse.pymc.io/t/can-we-access-the-component-distributions-in-a-mixture-model/3361/2 "2019-05-31T05:14:29Z")

</div>

By sample you mean posterior samples? In that case it is not possible as in a marginalized mixture model they are blend together.

---

<div class="post-metadata">

**Author:** ![rpgoldman](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/rpgoldman/32/1800_2.png) [@rpgoldman](https://discourse.pymc.io/u/rpgoldman)\
**Post date:** [June 4, 2019, 11:27pm UTC](https://discourse.pymc.io/t/can-we-access-the-component-distributions-in-a-mixture-model/3361/3 "2019-06-04T23:27:08Z")

</div>

Thanks. I see. In your schizophrenia example you make a mixture model “by hand,” and I think in there you can “look into” the mixture components and the mixing weights, because they are all normal PyMC3 entities, but in the `MixtureModel` the are just `.dist` objects.