Hello everybody,

I have a simple question about how to use the option unit_root_vars in estimation.

As far as I understand, it should be used when the observable has a unit root in the data. Does the observable in the model should also have a unit root, meaning that the model displays such an equation: obs=obs(-1)+err ?

Thank you in advance,


you have to declare as unit_root_vars ALL the unit root variables IN your model. In the Kalman filter, they will be initialized with a diffuse prior as in Durbin and Koopman. Check for updates at the end of the week: I just found some bugs with the diffuse prior.



Thank you , Michel.