I am reading DYNARE USER GUIDE (An introduction to the solution & estimation of DSGE model) Griffoli (2007-2008).
On page 56, Chapter 5. Estimating DSGE models - Basics, there is a TIPS indicates that
’‘If the plotted moments are unstable or do not converge, you may have a problem of poor priors’’
This is one of two suggestions for MH convergence problems. So I would have a question that
what the meaning the author mentions a problem of poor priors? Poor prior means that prior variance is too small or poor prior means a bad prior distribution?
Please correct my understanding
Thank you so much
I don’t think that statement in the user guide makes sense. If you have convergence issues, changing your prior is clearly not the first thing you should do.
Yes, this statement in DYNARE USER GUIDE makes me confusing as well.
It is also indicate later that
’‘Another approach is to undertake a greater number of MH simulations’’
So, when I have the MH convergence problem, first thing I do is that I increase number of interation for each Chain of MH. My approach is reasonable? if not, please correct me
First check whether your estimation starts at the mode, i.e. look at the mode_check-plots. Then look at the trace_plots to see why the MCMC has not converged. How many draws did you use?
Ah ok Prof. @jpfeifer
I am using Dynare 4.4 and your ‘’ An introduction to Grahps in Dynare’’ for old version of DYNARE
I will try to use the lastest version
If I have something unclearly, then I would let you know
Thank you so much for that