# Chains sampling completly different area of parameter space and never converge

**URL:** <https://discourse.mc-stan.org/t/chains-sampling-completly-different-area-of-parameter-space-and-never-converge/4885>\
**Category:** Modeling\
**Created:** [July 16, 2018, 3:07pm UTC](https://discourse.mc-stan.org/t/chains-sampling-completly-different-area-of-parameter-space-and-never-converge/4885 "2018-07-16T15:07:27Z")\
**Posts on this page:** 1\
**Showing post:** 7

<div class="post-metadata">

**Author:** ![yuling](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.mc-stan.org/yuling/32/13044_2.png) [@yuling](https://discourse.mc-stan.org/u/yuling)\
**Post date:** [October 8, 2018, 9:45pm UTC](https://discourse.mc-stan.org/t/chains-sampling-completly-different-area-of-parameter-space-and-never-converge/4885/7 "2018-10-08T21:45:55Z")

</div>

From the first graph, the “different regions”, or more accurately meta-stability, clearly comes from q2=0 and q3=0 I suppose both are sd parameters and it just a funnel-type. I would imagine the easiest solution is to try zap ([Zaps to avoid the funnel](https://discourse.mc-stan.org/t/zaps-to-avoid-the-funnel/5829)).

---

_[View the full topic](https://discourse.mc-stan.org/t/chains-sampling-completly-different-area-of-parameter-space-and-never-converge/4885)._