It's potentially misleading to run a shared model if it has samples already in it from individual model mcmc. Hierarchical model crashes if the models have samples. I think in both of these cases either it should error out if there are pre-existing samples, or it should clear the samples with a Warning.
Adding: also, should check that the domain of the parameters is consistent across the models (t and x orig_min and orig_max) and error out if not.
Adding: it would also be nice if the models could be passed in as a dictionary, in addition to as a list, maintaining the names of the models for subsequent query for parameters.
It's potentially misleading to run a shared model if it has samples already in it from individual model mcmc. Hierarchical model crashes if the models have samples. I think in both of these cases either it should error out if there are pre-existing samples, or it should clear the samples with a Warning.
Adding: also, should check that the domain of the parameters is consistent across the models (t and x orig_min and orig_max) and error out if not.
Adding: it would also be nice if the models could be passed in as a dictionary, in addition to as a list, maintaining the names of the models for subsequent query for parameters.