Hi, I’m also very new to pyMC, so take my input with a big grain of salt.
I think you may have issues inputing pyMC3 variables (nas,nbs) directly into your function.
In addition, if your model returns a vector, func should perhaps be multivariate.
In all cases, I’m not sure you can avoir using a black-box likelihood here, but I may be wrong. There is an example notebook here : Using a “black box” likelihood function — PyMC3 3.11.5 documentation
I hope this helps