The ability of the model checking procedure to detect model misfit
was also influenced by the discrepancy between the model and the
simulated data, as measured by the θ value used to simulate the data
(Fig. 2). In the low and middle noise cases (Fig. 2A and B), model misfit
could be detected for values of θ larger than or equal to 1.5, since pmin
values were sharply smaller than those obtained with the reference
model (with θ equal to 1). Model misfit could not be detected in the
large noise case, irrespective of the value of θ (Fig. 2C).