# Initializing Simplex / Dirichlet

**URL:** <https://discourse.mc-stan.org/t/initializing-simplex-dirichlet/6289>\
**Category:** Modeling\
**Tags:** performance, fitting-issues\
**Created:** [October 31, 2018, 3:34pm UTC](https://discourse.mc-stan.org/t/initializing-simplex-dirichlet/6289 "2018-10-31T15:34:46Z")\
**Posts on this page:** 1\
**Showing post:** 4

<div class="post-metadata">

**Author:** ![hhau](https://avatars.discourse-cdn.com/v4/letter/h/2bfe46/32.png) [@hhau](https://discourse.mc-stan.org/u/hhau)\
**Post date:** [October 31, 2018, 6:49pm UTC](https://discourse.mc-stan.org/t/initializing-simplex-dirichlet/6289/4 "2018-10-31T18:49:37Z")

</div>

> [@rtdew1](#):
>
> Thanks for your reply, you’re absolutely right about rounding inducing zeros, which yields the problem. I feel stupid for missing this.

Definitely don’t feel silly about this, it’s a subtle thing! - I only thought of this because I spent quite some time figuring out why [this other example](https://discourse.mc-stan.org/t/error-occurred-during-calling-the-sampler-sampling-not-done/6061/7) didn’t work.

> [@rtdew1](#):
>
> My intent is to use these simpler submodels to initialize parameters for the bigger model. One of those parameters is simplex valued.

Seems reasonable, I have done this before. You will loose the ability to diagnose a misbehaving model by its sensitivity to initial conditions, and the R-hat diagnostic doesn’t make as much sense. But sometimes it can be a way of getting sensible answers out. Good luck with it!

---

_[View the full topic](https://discourse.mc-stan.org/t/initializing-simplex-dirichlet/6289)._