Understanding mcmc_pairs output when nuts_params is passed through the np argument


I am using mcmc_pairs from bayes plot to understand cmdstanR output. In particular I pass the nuts_params function applied to the output of the sample function so that this function highlights divergences. However, beyond divergences I have realized that it also plots some gray dots which I can find what they mean in the docs. I have also not seen anyone having this problem. Here is a picture of the output and code to generate it:

nuts_params <- nuts_params(fit)
mcmc_pairs(posterior_draws, np = nuts_params, regex_pars = c('sigma','tau'))

Thanks in advance.

Near the bottom of the function reference page for mcmc_pairs just above the references, it says that points are colored differently for divergent transitions, reaching max-tree depth, or whatever statistic you’re looking for from the nuts_params() function. There’s even mention of a pairs_style_np() function that can be used to help style them.

1 Like