Can Dynare do bayesian estimation for non linear models? I thought the kalman filter can only handle linear cases with gaussian shocks?
Definitely, it could.
This is interesting because normally Kalman filter can’t handle non linear cases.
There must be some linearization done. ie. The “extended” Kalman filter is used, rather than the standard Kalman filter.
I couldn’t find the description of the exact non linear kalman filter algorithm in dynare manual.
Currently, Dynare only supports estimation of linearized models. Future version will contain particle filters and thus allow non-linear estimation.