Determinacy & prior simulate puzzle

Estimation ML/Bayesian estimation
1

Hi all,

I’m estimating a model with a relatively small determinacy region. I’ve run into what seems to be a puzzle though. When I plot the determinacy region, I can see that my priors do place mass in the indeterminacy region. However, when I use the “prior simulate;” command, it gives me a prior mass of 1 with no prior mass in the indeterminacy region.

What am I missing? Files attached.

With thanks,
DCF

P.S. Pardon the granularity in the determinacy plot; wanted to decrease run-time for this post.

upload.zip (38.4 KB)

2

This is a bug in the prior simulate function. See https://github.com/DynareTeam/dynare/issues/1607. Thanks for reporting.

3

Hi Prof Pfeifer,

Thanks for your reply! No problem on reporting.

Are you aware of any other method to calculate the prior mass falling into the indeterminacy region?

With thanks,
DCF

4

For a quick fix, go to prior_sampler.m and change the line

    [dr,INFO,M_,options_,oo_] = resol(work,M_,options_,oo_);

to

    [dr,INFO,M_,options_,oo_] = resol(0,M_,options_,oo_);

After that, it should work correctly.

5

Thanks, that does seem to be working.