search

alashworth / test-issue-import

0 stars 0 forks source link

propto control in sampling statements #27

Open alashworth opened 5 years ago

alashworth commented 5 years ago

Issue by bob-carpenter Friday Jan 17, 2014 at 12:20 GMT Originally opened as https://github.com/stan-dev/stan/issues/496

We should allow users to control whether constants are dropped in the model. Right now,

y ~ foo(theta1,...,thetaN);

will drop any terms in log foo(y | theta1,...,thetaN) that are constant, whereas

increment_log_prob(foo_log(y, theta1,...,thetaN))

will not. We could add controls so that either one can be used either way. It should be simple to implement because it's just a boolean template parameter to the same underlying function. I'm not sure what the syntax should be, but maybe something as simple as foo_propto_log() for the functions and foo_full() for the sampling statements. (It's unfortunate that the defaults are different, but I'd hate to break backward compatibility by changing that at this point.)

alashworth commented 5 years ago

Comment by bob-carpenter Thursday Nov 24, 2016 at 05:27 GMT

See #1299 for access in function definitions.

alashworth commented 5 years ago

Comment by syclik Wednesday Nov 30, 2016 at 14:56 GMT

I'm bumping milestones up on this issue. Since we've introduced target +=, it makes sense that we want to have this control on our functions.