Understanding the cause of divergent transitions (no apparent funnel behavior)

[bbbales2]

@martinmodrak

Sorry, short on time

Yeah same. I’ve been looking at this awhile and just haven’t responded. This is a good question but hard to respond to.

@tomli

In fact, the flagged transitions (red in the attached figure) seem to be distributed in proportion to the density.

This is the thread to read: Getting the location + gradients of divergences (not of iteration starting points)

But the short story is sometimes it’s just hard to get anything out of these plots.

@tomli

I’ve not included the code because the full code has many additional parameters that has been turned off in the example posted here. Happy to post it if someone’s interested and/or think it helps to understand the behavior.

We’re in guess and check mode. Do you mind posting the model? Numeric things can lead to divergences. Not doing things on the log scale, for instance (having any explicit exps and such).