Estimation/ imposing initial conditions on state variables

[Mag]

In Dynare estimation, is there a way to impose initial conditions on the state variables? Some elements of context on why we want to do such a thing (cause in principle, why not let the estimation pick what gives the best fit?). We are estimating a model with expectations deanchoring. Long-term inflation expectations play a central role in the model so we really want to match the data (the michigan 5-year-ahead expectations) on this dimension. So we estimate the model in two steps. In a first step we estimate the parameters related to deanchoring (essentially a learning gain) using inflation and 5-year-ahead expectations as observables. In a second step we estimate everything else using all observables. It works, but we realize the second-step estimation assigns slightly different values to the initial condition of long-term inflation expectations (a state variable of the model). This makes sense (it is now chosen to fit many more things), but ideally we would like to avoid this. Is there a way to impose the initial conditions of the state variables in this second step?
Thanks for your help

[jpfeifer]

See whether

Filter_initial_state Block

By default, the Kalman filter is initialized at the steady state with the unconditional variance of the states. With the filter_initial_state you can alter the mean of this initialization. We do not have a interface for altering the variance. But that does not fix the smoother mean for the first period to a particular value. First of all, we are talking about the initial value before the first observation. Second, the smoother does a forward and a backward pass. The latter will also alter out b…

is what you are looking for.

[Mag]

Thanks a lot. It is very useful !
Best