number of shocks and number of observed variables

[Dave]

Dear all,

Dynare requests that the number of observed variables must be smaller or equal to the number of shocks. Suppose the model that I am going to estimate has the same no. of shocks as the no. of observed variables.If I want to shut down one or two shocks and re-estimate the model, I have more observed variables than the shocks in the model. How can I resolve this problem? Thank you.

Regards,
Dave

[StephaneAdjemian]

Hi Dave,

Dynare requests that the number of shocks must be greater or equal to the number of observed variables because otherwise the likelihood is not defined (there is a stochastic singularity and the covariance matrix of the prediction errors is not full rank). If in your model you have less structural shocks than observed variables you can:

1. Add measurement errors.
2. Remove some of the observed variables.
3. Consider an empirical strategy not relying on the likelihood, for instance simulated moments or indirect inference (these methods are not yet implemented in dynare so you have to code it yourself).

Best, Stéphane

[Dave]

Dear Stéphane,

Thank you for your reply.

Best,
Dave