Hi, I fit a model hierarchically, sampling priors for all three parameters for all subjects from the same set of priors. I end up getting one waic score per model for all subjects for model comparison. I also hope to compare models for each individual subject. Is there a way for me to do it without having to fit the model for each individual subject again? For example, maybe separate the model trace output (which is an arviz object) into one for each subject?
with pm.Model() as m: alpha = pm.Beta('alpha', alpha=1, beta=1, shape=n_subj) beta = pm.Gamma('beta', alpha=3, beta=1/2, shape=n_subj) decay = pm.Beta('decay', alpha=2, beta=15, shape=n_subj) param = at.as_tensor_variable([alpha, beta, decay], dtype='float64') like = pm.DensityDist('like', param, logp=agent.aesara_llik_td,observed=data_vec)