Fix prior sampler
prior simulate
in cli/prior.m
calls
results = prior_sampler(0, Model, BayesOptions, options_, oo_, EstimatedParams);
where the 0 indicates that no decision rules need to be saved. The problem is that this flag drsave
is then used to call resol
with
work = ~drsave;
[dr,INFO,M_,options_,oo_] = resol(work,M_,options_,oo_);
so that resol
only computes eigenvalue, but not BK conditions etc. As a result, the reported simulations will never fail the BK conditions. See https://forum.dynare.org/t/determinacy-prior-simulate-puzzle/11681