We currently have e.g.
likelihood_dict = {
"type": "gaussian",
"prec_o": 4,
"prior_hyperparameters": {"type": "gamma", "alpha": 2.0, "beta": 2.0},
or
"phi": 0.5, # has to be between 0 and 1
"tau": 3, # precision
"ph_phi": {"type": "beta", "alpha": 5.0, "beta": 1.0},
"ph_tau": {"type": "gaussian", "mean": 0.0, "precision": 0.5},
where phi clearly belongs to ph_phi and tau to ph_tau. I think it would be good to have an argument in ph_phi that just passes the name phi explicitly so that its clear that they belong together. This will be more relevant if we have two priors of the same type or name where it might otherwise be confusing and it also makes it structurally more clear that they belong together. In the first example its already not so clear that prec_o belongs to prior_hyperparameters. We can also consider making it more clear that those are initial guesses.
We currently have e.g.
or
where phi clearly belongs to ph_phi and tau to ph_tau. I think it would be good to have an argument in ph_phi that just passes the name phi explicitly so that its clear that they belong together. This will be more relevant if we have two priors of the same type or name where it might otherwise be confusing and it also makes it structurally more clear that they belong together. In the first example its already not so clear that prec_o belongs to prior_hyperparameters. We can also consider making it more clear that those are initial guesses.