Bridgestan handling param constraints

Modeling
1

Hi,

I’m using the Bridgestan (2.6.1) log_density() function on the eight schools model in posteriordb, which includes a tau param:

...
  real<lower=0> tau;
... 
}

I should be getting -inf for any draw with a negative tau (right?), but I’m not. Bridgestan seems to be ignoring the constraint on the parameter. Is this a bug?

Thanks!

2

BridgeStan expects the inputs on the unconstrained scale. If you have values on the constrained scale you want to use, there is an unconstraining function provided