Bayesian estimation: Orthogonalized shocks


#1

Dear Dynare community,

I’m doing an stochastic estimation by means of Dynare. It works perfectly; nevertheless the results of orthogonalized shocks to e_z are the next:

6_orthogonalized_shock_to_e_z

What do you think about it? Are consistent or it could be displaying an error on the estimated parameters?

Thank you very much in advance,
Best regards!


#2

What is your worry? The immediate issue is that you are using an old Dynare version where the graphs are not displayed properly if they cross the 0 line.


#3

Dear @jpfeifer ,

Yes, maybe my perception is the issue with the graphs in old dynare version. Currently I’m using dynare 4.4.3 because newest versions reported me issues in the Kalmann filter related with stochastic singularity…

I already tried to use calibration as starting values and to restructure the shocks, but i could not find the mistake that generates the error…

Find attached mod. file and xlsx observed data in case you want check the model.

Many thanks in advance

macrofish_discards_estimation.mod (2.5 KB)

L_D_data_2004_2015_test.xlsx (9.9 KB)


#4

Stochastic singularity is a fundamental problem with the estimation setup. Your model implies an exact linear combination between the observables. Using an older Dynare version that did not check for this issue is not a solution here. One way of debugging is to drop one observable at a time and see when the problem disappears.