Target variables are functions of Gaussians (aka how to define a deterministic multivariate distribution?)

Is f(a,b) a CDF? (as a PDF it doesn’t necessarily need to be in [0,1])?
Why is b not defined outside loop, as it doesn’t depend on observations?

I think @dlakelan solution might work, just adding:

@addlogprob!(log(f(observation[i]; a=a, b=b))

instead of using observations[i].

Does this sound resonable?

2 Likes