SBC

[teddygroves]

Add a step implementing model checking by simulation based calibration

[mitzimorris]

a general procedure for SBC in and of itself would be huge.
the stumbling block is the procedure for generating data.
for a given Stan model, the data-generating process is somehow implicit in the model.
but to test with SBC, we need to be able to write the script or program to generate the simulated data and then do the simulations, which means running the model 100s of times. here's where clusters can help, but still, hugely expensive computationally.