How to perform weighted inference

Questions
3

Thank you for the prompt answer!

I’m not very familiar with the framework but Potential doesn’t allow to input observed values. Anyhow I’ve recycled you idea using DensityDist and I could make it work:

def weighted_logp(w,distributuion,**kwargs):
    return distributuion.dist(**kwargs).logp

model = pm.Model()
with model:

    b0=pm.Normal('b0',mu=0,sd=.1)
    b1=pm.Normal('b1',mu=0,sd=.1)
    mu=pm.math.exp(b0*X[:,0]+b1*X[:,1])
    y_w=pm.DensityDist('Y_w',weighted_logp(w,pm.Poisson,mu=mu) , observed=y)

    trace=pm.sample(1000)

Thanks again!