After "prior simulate" What is the link between modelname_results.mat and endogenous variables

Dear all,
To investigate problems in a model, it was suggested that one give the two commands
estimation (where you put mode_compute=0, mh_replic=1, order=1 and provide a dataset)
prior simulate;
The model is good if the (effective) prior mass is close to 1.
Below is what I found for my model:

negative values are not allowed

Prior mass = 0.023
BK indeterminacy share = 0
BK unstability share = 0
BK singularity share = 0
Complex jacobian share = 0
mjdgges crash share = 0
Steady state problem share = 0
Complex steady state share = 0
Analytical steady state problem share = 0.977

Total computing time : 4h07m19s
Note: warning(s) encountered in MATLAB/Octave code


I have loaded the 4x4 matrix “modelname_results.mat” generated after estimation

load(‘January2018cmrV3_results.mat’)
view

ans =

1.0000         0         0   -0.5000
     0    1.0000         0   -0.5000
     0         0   -1.0000    9.1603
     0         0         0    1.0000

My questions are:

  1. Is there any link between the values of the elements of that results matrix and my endogenous variables?
  2. I have noticed that the risk premium term coefficients are very influential in the model. If I calibrate one of them the “negative value not allowed” error shows at the Identification step. If I estimate both parameters, the “negative value not allowed” error shows later in the Estimation step. Is there a way to see which of the endogenous turns negative in the estimation process to help fix the model?
  3. Suppressing altogether the shock on the risk premium give “the matrix must be positive definite” message error.
    I would really appreciate if someone could give a hint in all those problems, may be even proceed to have a fresh look at my model because I feel like I am half way home.

Thank you for your time.

Does this mean you problem is solved and I should delete this post?

Hello jpfeifer,
Thank you very much for picking my question once more.
Unfortunately, I have not solved the problem. As I was tweaking the value of one parameter (ubar) to improve the prior mass, an identification error crept into the code and I thought it should be the cause my problem. I have solved the identification problem while maintaining the improved prior mass but at least one of my endogenous variables is still negative.

“negative values are not allowed”

Prior mass = 0.35555
BK indeterminacy share = 0
BK unstability share = 0
BK singularity share = 0
Complex jacobian share = 0
mjdgges crash share = 0
Steady state problem share = 0
Complex steady state share = 0
Analytical steady state problem share = 0.64445
At this stage, I wonder why all the above error sources are zero but the mode of the model is still unobtainable. My questions are as follows:

  1. Are the priors to be changed?
  2. Should I change the values of calibrated parameters?
  3. Is the “analytical steady state problem share” relates to a fundamental flaw in the model? In that case, is the 4 x 4 results matrix below contains any clue for the troublesome endogenous variable (s)?

load(‘January2018cmrV3_results.mat’)
view

ans =

1.0000         0         0   -0.5000
     0    1.0000         0   -0.5000
     0         0   -1.0000    9.1603
     0         0         0    1.0000

January2018cmrV3.zip (24.3 KB)

I would really appreciate if you take the time to look into these questions.

You need to be more systematic here. It is very hard to follow i) what you are trying to do, ii) which type of problem arises, iii) what you already tried to debug the problem, i.e. which clues you already have.

Dear jpfeifer,

I misunderstood the meaning of “Analytical steady state problem share” . “Analytical SS problem” can happen even though all previous diagnostics after “prior simulate” are OK because it is also a share of estimation problems. Thank you very much for your patience and your insight. I have solved my problem and the code is running.