The RATS Software Forum

Dear Tom:
In ch 9.3 Non-linearities in the Equations, RATS Handbook for State-Space Models, 2e,
The said that “The difficulty there is that the smoothed estimates change every time the expansion points change, even if the model itself is treated as known, which requires either iterating to convergence or doing a single smooth at reasonable guess values and sticking with it.”

In @disaggregate and example of “Employing the extended Kalman filter in measuring the output gap”, it uses iterating. In example of “The Great Recession and the inflation puzzle”, it uses a single smooth. How can we select strategy for dealing with EKF? Is there rule or empiric?


Best Regard
Hardmann

hardmann wrote:Dear Tom:
In ch 9.3 Non-linearities in the Equations, RATS Handbook for State-Space Models, 2e,
The said that “The difficulty there is that the smoothed estimates change every time the expansion points change, even if the model itself is treated as known, which requires either iterating to convergence or doing a single smooth at reasonable guess values and sticking with it.”

In @disaggregate and example of “Employing the extended Kalman filter in measuring the output gap”, it uses iterating. In example of “The Great Recession and the inflation puzzle”, it uses a single smooth. How can we select strategy for dealing with EKF? Is there rule or empiric?


Best Regard
Hardmann

Not that I'm aware. Note that EKF (just like the original KF) originated in engineering/hard sciences where the need for smoothing is fairly rare---you have to control processes in real-time, so the filter is the tool, so there is much less of a literature on smoothing with non-linear models.