The Hessian is negative

This forum is closed. You can read the posts but cannot write. We have migrated the forum to a new location where you will have to reset your password.
Forum rules
This forum is closed. You can read the posts but cannot write. We have migrated the forum to a new location (https://forum.dynare.org) where you will have to reset your password.

The Hessian is negative

Postby phuong.macro » Fri Jun 30, 2017 10:12 am

Hi all

after finding the global mode with the mode_compute = 7 (many number of interation and number of fuction evaluation as well) then the likelihood of -2197.2944 can not be improved anymore, Dynare also report that the Hessian matrix is negative. So that MH algorithm can not be computed.
I try with another mode_compute, for example, 4 and 9, but the Hessian is still negative in Dynare and the likelihood of -2197.2944 can not be improved anymore either
Anyone has a good suggestion for me?
Thank a lot
phuong.macro
 
Posts: 28
Joined: Mon Jun 27, 2016 7:35 pm

Re: The Hessian is negative

Postby Robert » Fri Jun 30, 2017 10:37 am

What happens if you try mode_compute = 6? This one is Monte-Carlo based. Another option would be that you load your last mode file and keep searching for the posterior mode. Use the mode_file = yourfile option in the estimation command.
Robert
 
Posts: 36
Joined: Wed Jan 13, 2016 3:10 pm

Re: The Hessian is negative

Postby phuong.macro » Fri Jun 30, 2017 12:23 pm

Hello Robert

Thank you for your suggestion
But I did it already before asking for help (mode_file=ALEstim_phuong_mode) and try to use another method_mode, for example 9 which seem to be the best mode finder. But the hessian matrix is still negative
phuong.macro
 
Posts: 28
Joined: Mon Jun 27, 2016 7:35 pm

Re: The Hessian is negative

Postby jpfeifer » Fri Jun 30, 2017 12:27 pm

What do the mode_check plots say where the problem is?
------------
Johannes Pfeifer
University of Cologne
https://sites.google.com/site/pfeiferecon/
jpfeifer
 
Posts: 6940
Joined: Sun Feb 21, 2010 4:02 pm
Location: Cologne, Germany

Re: The Hessian is negative

Postby phuong.macro » Fri Jun 30, 2017 1:35 pm

Dear Prof. Pfeifer

I wound to attach all my files a long with data. It is convenient for me to describe what is going on with my model
Attachments
Trans_Aus_Dat_dynare.xls
data file
(45.5 KiB) Downloaded 58 times
ForFisEstRes_Pmod.m
matlab file
(3.36 KiB) Downloaded 57 times
ALEstim_phuong_steadystate.m
steadystate file
(22.83 KiB) Downloaded 56 times
ALEstim_phuong.mod
mod file
(31.02 KiB) Downloaded 61 times
phuong.macro
 
Posts: 28
Joined: Mon Jun 27, 2016 7:35 pm

Re: The Hessian is negative

Postby michall23full » Sat Jul 01, 2017 3:57 pm

I recommend to begin with mode_compute=6 (it always results in positive definite Hessian) and then try another mode-finder e.g. 9 providing result file from 6 as mode_file.
michall23full
 
Posts: 12
Joined: Thu Jun 08, 2017 10:45 am

Re: The Hessian is negative

Postby jpfeifer » Sun Jul 02, 2017 10:15 am

You did not provide the mode-file. Please upload everything in one zip-file
------------
Johannes Pfeifer
University of Cologne
https://sites.google.com/site/pfeiferecon/
jpfeifer
 
Posts: 6940
Joined: Sun Feb 21, 2010 4:02 pm
Location: Cologne, Germany

Re: The Hessian is negative

Postby phuong.macro » Mon Jul 03, 2017 9:47 am

Dear Prof. Pfeifer

I would attach all file as zip file which include the mode_file
Attachments
Question.rar
(42.08 KiB) Downloaded 60 times
phuong.macro
 
Posts: 28
Joined: Mon Jun 27, 2016 7:35 pm

Re: The Hessian is negative

Postby phuong.macro » Mon Jul 03, 2017 9:51 am

Dear michall23full

but mode_compute= 6, It's not an optimization routine!... Or a very inefficient one
phuong.macro
 
Posts: 28
Joined: Mon Jun 27, 2016 7:35 pm

Re: The Hessian is negative

Postby jpfeifer » Mon Jul 03, 2017 2:15 pm

You can clearly see that your "mode" is not at the maximum for b and rdeltay (which are also far away from the prior mode). Try to understand why the model/data push the estimates to that region.
------------
Johannes Pfeifer
University of Cologne
https://sites.google.com/site/pfeiferecon/
jpfeifer
 
Posts: 6940
Joined: Sun Feb 21, 2010 4:02 pm
Location: Cologne, Germany

Re: The Hessian is negative

Postby phuong.macro » Wed Jul 12, 2017 7:53 am

Yes thank you Prof.Pfeifer
phuong.macro
 
Posts: 28
Joined: Mon Jun 27, 2016 7:35 pm


Return to Dynare help

Who is online

Users browsing this forum: Google [Bot] and 11 guests