# Simple GP model giving large MAP values for covariance timescale

**URL:** <https://discourse.pymc.io/t/simple-gp-model-giving-large-map-values-for-covariance-timescale/7844>\
**Category:** Questions\
**Created:** [August 4, 2021, 11:17pm UTC](https://discourse.pymc.io/t/simple-gp-model-giving-large-map-values-for-covariance-timescale/7844 "2021-08-04T23:17:01Z")\
**Posts on this page:** 1\
**Showing post:** 3

<div class="post-metadata">

**Author:** ![jlindbloom](https://yyz2.discourse-cdn.com/flex036/user_avatar/discourse.pymc.io/jlindbloom/32/5016_2.png) [@jlindbloom](https://discourse.pymc.io/u/jlindbloom)\
**Post date:** [August 5, 2021, 2:00am UTC](https://discourse.pymc.io/t/simple-gp-model-giving-large-map-values-for-covariance-timescale/7844/3 "2021-08-05T02:00:13Z")

</div>

I messed around with your code and found the same thing. I have a feeling that you’re experiencing the same issue that Michael Betancourt talks about in [this blog post](https://betanalpha.github.io/assets/case_studies/gp_part3/part3.html#2_inferring_the_hyperparameters_with_a_naive_bayesian_fit). See [this thread](https://discourse.pymc.io/t/how-to-put-a-restriction-on-the-value-of-hyper-parameters-in-gaussian-process/6072/2) for some advice. Long story short, priors are important (but especially with GPs)!

---

_[View the full topic](https://discourse.pymc.io/t/simple-gp-model-giving-large-map-values-for-covariance-timescale/7844)._