Ragged array model less efficient, converges with different lp__ than non-ragged version?

[David Manheim]

I'm using the same data in both models, and get (almost) the same results - but the lp__ is very different.

I understand why the ragged version is slower; I'm confused by the different lp__.

The attached are the models, data, and script to run them in R. (It should work out of the box if they are all in the current R directory.)

[David Manheim]

Based on the conclusion of a twitter discussion with Krzysztof Sakrejda...

I'm silly - I excluded part of the model in the ragged version.

Nevermind.