Likelihood function in the scan function (AR2 model)

7

Let me ask one more question.
What should I do when I have several different likelihoods, for example?

        y_t = pm.Normal.dist(mu=m_t, sigma=sigma)
        y_p = pm.Poisson(mu=m_t)

Do I just give a (num_data, 2)-shaped matrix with normal distribution y and Poisson distribution y as the observed argument?

   ar_innov = pm.CustomDist(
        name,
        params,
        observed=y_obs,
        shape=(n_steps,)
    )