Pareto K diagnostics and kfold model comparison

[sbashevkin]

@paul.buerkner

@sbashevkin Please use the “Interfaces - brms” tag for brms related questions. Otherwise, I will likely overlook it (I changed the tag manually this time).

Sorry about that. I thought this question would be a more general one about interpretation of loo diagnostics.

@paul.buerkner

With regard to the number of parameters, parnames() won’t help you. To what kind of number of parameters are you referring to? Effective number of parameters?

I am interested in the number of parameters that I should be comparing to the p_loo estimates, as @avehtari refers to here: A quick note what I infer from p_loo and Pareto k values - #3 by avehtari where he says

@avehtari

start by counting all parameters in the parameters block, and then refine if necessary.

I was thinking that the results of parnames() would be equivalent to the number of parameters in the Stan parameters block?