The question about "Warning: Matrix is close to singular or badly scaled. Results may be inaccurate."

Dear professors:
I am sorry to disturb you. when I run the Dynare, I get the following message: “Warning: Matrix is close to singular or badly scaled. Results may be inaccurate”. However, I can get the estimation results without any errors. I have searched related topics. Some said it can be ignored, while others advise changing the initial value of parameters. Thus, my questions are:

  1. Under what conditions can the warning be ignored?
  2. How can I set the “proper” initial value?
  3. By the way, when I update Dynare to 5.0, I get the message “Installation error The installer encountered an error that caused the installation to fail. Contact the software manufacturer for assistance.” Why I cannot update Dynare above 4.6.4 on my Macbook? Is it because I haven’t updated the operating system?
    Best wishes!
    resiliencet.mod (8.2 KB)
    all.mat (2.2 KB)
  1. It very much depends on where that message appears. In your case it happens when trying to compute the unconditional variance of the states for the Kalman filter. That is rather unusual. Why do you specify diffuse_filter although the model does not have unit roots? Have you tried whether the problem also appears in a simpler version of your model?
  2. Each Dynare version is compatible with a particular macOS unless you build it from source. But without the log-file it is impossible to tell why the installation failed.

Dear Prof. Jpfeifer:
Thanks for your replying.

  1. Do you mean I should simplify my model to find out the reason? Could you please give me some possible reasons and how to solve these problems?
  2. After viewing the log-file, I find that it is not able to connect to I have updated to Dynare 5.2 by skipping to install use_dll option. How does use_dll option affect the operation of Dynare?

Best wishes!

  1. Yes. I don’t know what causes the issue. That’s why I suggested to simplify the model.
  2. use_dll is an option to model(use_dll) to trigger the use of precompiled files. It increases the run-time speed.

Dear Prof. Jpfeifer:
Thanks for your advice. After deleting some parameters in the estimated_params block, I can run the Dynare without warning.

  1. Is it a problem about the parameter setting?
  2. How to set parameters correctly?
  3. What literatures would you recommend?

I found that I sometimes even got the error with incorrect parameter settings.
Best wishes!

  1. That is impossible to tell, but seem likely if the problem goes away with different values.
  2. There are many way to parameterize the model. Most often, people simply follow the literature.
  3. That is also impossible to tell. Usually, you have a look at the most closely related papers and follow their lead.