The current implementation of the generated quantities functions allows the users to provide a list of patients and a time grid as 2 independent arguments. The functions then predict the values for all patients at the given time grid.
Unfortunately this doesn't quite meet two common use cases which would be:
Predict just at the observed time points for each patient (which is different per patient)
Predict at say X evenly spaced intervals between 0 and the patients event time
I think we potentially need 3 changes
[x] Generalise the looping mechanism in the generated quantities to avoid each model having to re-implement the same logic
[ ] Allow user arguments to generated quantities to specify different time points per subject
[ ] Have helper objects / functions to setup the per patient grids as specified above
The current implementation of the generated quantities functions allows the users to provide a list of patients and a time grid as 2 independent arguments. The functions then predict the values for all patients at the given time grid.
Unfortunately this doesn't quite meet two common use cases which would be:
I think we potentially need 3 changes