taken from #299 (comment)
Maybe we could take a page from the plots side here and allow a stats dictionary argument. That being said, here I would only allow providing pre-computed elements to keep things simple. For example:
rhat = az.rhat_nested(idata, superchains=...)
az.diagnose(idata, stats={"rhat": rhat})
In that particular case it would make it easier to take advantage of diagnose even if wanting to use rhat_nested instead of regular rhat. They could also change the probability at which to compute ess_tail
taken from #299 (comment)
Maybe we could take a page from the plots side here and allow a
statsdictionary argument. That being said, here I would only allow providing pre-computed elements to keep things simple. For example:In that particular case it would make it easier to take advantage of
diagnoseeven if wanting to use rhat_nested instead of regular rhat. They could also change the probability at which to computeess_tail