Fix pareto k threshold typo in reloo function

[teddygroves]

Description

I noticed that changing the k_thresh argument to the function reloo from the default value 0.7 had no effect. I think this is because the threshold is accidentally hard-coded as 0.7 inside the function.

This change fixes the apparent typo, replacing 0.7 with k_thresh.

Checklist

• Follows official PR format
• Includes new or updated tests to cover the new feature
  This function is so far untested and I didn't think it made sense to write a whole test before fixing this bug.
• Code style correct (follows pylint and black guidelines)
• Changes are listed in changelog

[OriolAbril]

Thanks! It is great to see people using reloo, I wasn't really expecting it actually. I myself have not used it too intensively.

Let me know if you have any feedback on the sampling wrappers or on reloo itself (other than the calculation of p_loo being plain wrong). Would you find is helpful if for example reloo passed both label and positional indexes to the sampling wrapper sel_observations method? Are the docs on sampling wrappers clear?

[teddygroves]

No worries, thanks a lot for implementing this - it's really improving my modelling workflow.

I'm still getting used to the interface and don't have the expertise to help with the p_loo calculations but I'll certainly come back with feedback soon.

[OriolAbril]

I'll certainly come back with feedback soon.

Feel free to contact here on GitHub or at either of Stan or PyMC3 discourses.

[OriolAbril]

Thanks for the PR @teddygroves !