Hi,
I am estimating a model with Dynare. I need to add measurement errors to my observed variables, to model these measurement errors as AR(1) processes and set the variance of the measurement errors equal to 5% of total variance of the observed variables (I refer to McGrattan, Rogerson and Wright, 1997, IER, for this).
Here is what I do in the .mod file. After specifing the equations of the model, I add 2 measurement equations as follows:
oh=h+meh;
ou=u+meu;
where oh and ou are my observables while h and u are the corresponding variables of the model. meh and meu are the measurement errors.
Then I set the measurement errors to evolve as AR(1) processes:
meh=rhohmeh(-1)+hs;
meu=rhoumeu(-1)+us;
Then in the estimated_params section I write:
stderr us,0.1,0.01,3,INV_GAMMA_PDF,0.1,Inf;
stderr hs,0.1,0.01,3,INV_GAMMA_PDF,0.1,Inf;
stderr oh, 0.0005,0.0001,0.00199,INV_GAMMA_PDF,0.0005,Inf;
stderr ou, 0.0005,0.0001,0.002199,INV_GAMMA_PDF,0.0005,Inf;
Is this right? it is consistent with what I would like to do? should all these standard errors be estimated?
because the code has some difficulties to estimate the posterior mode.
Thank you in advance
Alice