Ah okay, interesting - so does your data look like? :
| participant_id | trial_id | x_val | y_val |
|---|---|---|---|
| 0 | 0 | 2 | 3 |
| 0 | 1 | 2.5 | 3.5 |
| 0 | 2 | 2 | 3.2 |
| 1 | 0 | 7 | 3 |
| 1 | 1 | 6 | 4 |
(… skipping rows for trial_id → 59)
That indeed is not what the examples assume: they assume a single prior covariance matrix for the entire set of observations.
Can I assume that you want to calculate a prior covariance matrix for each participant, created from their sequence of trials? If so, you’d probably need to do some looping so as to get multiple LKJCholeskyCovs into your model.
But then again, are really you sure you want to estimate a different prior covariance matrix per participant? This would imply that Alice has a uniquely different covariance to Bob, and it wouldn’t let you generalise anything about these people such that you could make predictions about Carol. etc