Minor bug (?) in latest rstan (2.19.2)

Before I open an issue on github, just want to verify I’m not misunderstanding something.
I ran a model, and forgot 2 positivity constraints. It’s not identified, so Rhats were large.
When it finished, it said ‘the largest R-hat is 1.75’, but in summary(fit)$summary, I saw several > 7.

Does that warning pertain to Rhat /within/ a chain? Or is it based on the new R-hat?

1 Like