What to do when mode check plots are problematic

Estimation ML/Bayesian estimation
1

Hi everyone,

I have a model that can be estimated using mode_compute=6, but I get very problematic mode check plots and I do not know how to tackle this problem. I have multiple problems in my plots:

  1. I have many red dots, this does not surprise me or worry me too much, because the solution of the steady state does not allow a big range for many parameters.
  2. I have plots (e.g. alppha), where the log-posterior and the log-lik kernel are not smooth continuous and importantly do not cross the vertical line indicating the estimated mode.
  3. I have plots (e.g. xxi), where the lines approach the vertical line from one side and may be tangent to the vertical line, but then stop. This, if I understand correctly can come from a corner solution.
  4. I also have plots, where there are no lines at all, but a vertical line is still visible.
    imageimage
    image

I don’t know what I need to do to fix these problems. Does anyone have an idea as to how to improve the plots? I attached the mod, steady state file and data file also to this post (The model in Dynare was log-linearized).

EnvironmentA.mod (7.4 KB)
EnvironmentA_steadystate.m (8.3 KB)
MyData.xlsx (10.6 KB)

2

One thing you may want to do is improve your error handling in the steady state file:EnvironmentA_steadystate.m (8.5 KB)
But that will not change the mode_check-plots substantially. The issue is indeed that the steady state does not exist for many parameter combinations. As such, you will obtain a corner solution and the mode_check-plots are not very informative.

3

Thank you very much for your adjustments!