Confidence band of Bayesian IRF (SW2007)

[dk19]

Dear Dynare forum,

I ran Smets and Wouters (2007) model using their data obtained from AER website. The posterior estimation results were nice but the bayesian irf looked rather strange, especially the confidence band, (1) sometimes there is no band, just the upper bound, and sometimes the band showed white area, instead of normal grey color; (2) the graph drew horizontal line, but not at (0,0).

The command used is:
estimation(datafile=usmodel_data,mode_compute=6,first_obs=71,mh_replic=20000,mh_nblocks=2,mh_jscale=0.20,mh_drop=0.2,bayesian_irf);

Dynare version 4.3.3 is used.

Has anyone encountered similar problems before? Any idea on what's wrong or what I can do to make the irf looks normal, like additional options?
Thank you so much!

Best,
DK19

[dk19]

I have carried out many attempts like increasing the number of mh_replic or using another country's data, but Dynare kept delivering the strange-looking confidence bands for the Bayesian IRF.

I would really appreciate if someone would be so kind to comment or share your ideas.Please, any kind of suggestions/insights are highly valuable and welcome.

Thank you so much in advance!
DK19

[jpfeifer]

This is a bug. Please upgrade to Dynare 4.4.3 and then replace the original Dynare\matlab-folder file with the attached one.

[dk19]

Dear Professor Pfeifer,

Thank you so much for your reply! It's very very kind of you!!!

The strange confidence band has kept me and my advisor confused for quite a while. If it is a bug, then I am relieved now. I will upgrade the new version of Dynare and do according to your instruction.

Please have a pleasant day, sir!
DK19