Thanks for that.
The motivation for avoiding a single model (with or without hyperparameters) was to be able to deal with situations where you might have ~100 data points per participant, and possible 1000’s of participants. If each model has 3 parameters (for example) then are ~3000 parameter models feasible to estimate in PyMC3?
I’m happy to implement single models which fit all participants (solution A), just that for my analysis plan I’d want to look at both individual fits and collective fits, just it would be slightly simpler if I could do that with only one single-participant set of models. But it looks like Solution A is going to be the simplest for the moment… implement single participant models, and multiple participant equivalents (either hierarchical or not).