Interpret pareto k diagnostic

Thank you for your reply @Bob_Carpenter !

I had seen the recommendations on what to do when there are too many high Pareto k values in other documentation, and to look into p_loo for further information on the underlying problem. I had not see the loo moment matching though and am now looking into it!

Something else I have not found is how many high Pareto k values is too many. Is there any “rule of thumb”, something related to the number of observation? Is only one high Pareto k value too much and model need reworking?

I can’t seem to find anything regarding actual threshold of high Pareto k values.

Thank you again!