How to build transition and measurement equations

Estimation ML/Bayesian estimation
1

Dear Johannes Pfeifer,

I would like to build state transition and measurement equations from the decision rules that DYNARE produces. I am using Dynare 4.3.3. I have read a previous post in this forum Dimension of dr_.ghx awkard (zeros omitted?) that partially explained this topic. In particular, it suggested using the function kalman_transition_matrix(dr_) for the state transition equation. However, I believe that in the Dynare version 4.3.3 this output is missing. Moreover, there was no explanation at all about the measurement equation. Please could you provide me more information about this?

Moreover, reading your paper “Observation Equations” in Section 1.2 you state that “In case of no measurement error, the covariance matrix of epsilon_t_obs is just a 0 matrix”. Does this mean that in your paper eq. (17) becomes y_t_obs = h (x_t) (where y_t_obs are the observed variables and x_t the structural shocks)?

Thank you.

Kind regards,

Marco

2

  1. Please have a look at github.com/JohannesPfeifer/DSGE_mod/blob/master/FV_et_al_2007/FV_et_al_2007_ABCD.mod
    and
    github.com/JohannesPfeifer/DSGE_mod/blob/master/FV_et_al_2007/ABCD_test.m
    that shows how to use the kalman_transition_matrix to generate the ABCD matrices. Note also that we do not provide detailed support for older version. Please upgrade to 4.4.3.

  2. Yes, in that case y_t^obs=h(x_t), but where x_t is the state variables at time t, not the structural shocks. x_t captures both states at time t-1 and structural shocks occuring at time t.

3

Thank you very much Johannes! What you suggested me worked perfectly.

Kind Regards,

Marco