Dear Professor.
I am getting the following error when I run the code for Bayesian estimation:
initial_estimation_checks:: Estimation can’t take place because too many shocks have
been calibrated with a zero variance!
This error has been bothering me for a long time, and I sincerely look forward to your reply!
bayesian.mod (6.7 KB)
mydata.mat (3.2 KB)
You have 6 shocks and 4 observables. But only 3 shocks have a non-zero variance. The other were set to 0. But you need at least 4 shocks with positive variance to estimate the model.
Thank you very much for your answer, but I still have a few questions: first, how to see exactly which shocks have non-zero variance; second, why do I have different variance when my settings for the shocks are all of the same form?
And according to your answer, I removed 1 observable, but the following error was reported:
initial_estimation_checks:: The forecast error variance in the multivariate Kalman filter became singular.
bayesian.mod (6.6 KB)
thedata.mat (2.4 KB)
- You set
shocks;
var epsilon_IG;stderr 1;
%var epsilon_G;stderr 1;
var epsilon_tau;stderr -1;
var epsilon_xi;stderr 1;
%var epsilon_a;stderr 1;
%var epsilon_x ;stderr 1;
end;
which means you only set three non-negative variances.
2. Having as many shocks as observables is only a necessary condition. There may still be stochastic singularity.
Thank you very much, I understood what you meant and ran the results.
I ran a similar Bayesian estimation procedure at the same time, and the results of that code did not run either, showing: matrix close to singular values, or scaling error. I just made some detailed changes in the model, but the form of the shocks and the data collection process did not change. Can you help me to see what is wrong with this code?
bayesiannew.mod (10.3 KB)
thedata.mat (2.4 KB)
model_diagnostics suggests that your model still has a singularity problem.
I am a dynare newbie and after a long day of thinking still don’t understand how to solve the singularity problem. My current thinking is that I can solve the singularity problem after solving the Colinear equations problem, but model_diagnostics shows: Colinear equations 18 19 21 23, but these equations are a few classical equations in the vendor sector. The change I made was to add the government’s capital to the production function, and my equations for the vendor sector before log-linearization were as follows
The equations after log-linearization are as follows:
As you can see, this is where I introduce human capital h and government capital KG, but such an operation is not really uncommon, so why are there colinear equations in these equations?
Meanwhile, dear professor, I encountered too many errors in the process of running the model and would like to ask if you have any summary information for dynare errors?
It’s often the case that such a message is triggered if there is a redundant equation, i.e. all the N-1 equations imply the Nth one.