Re: [Gretl-devel] user access to the Kalman filter
On Wed, 15 Apr 2009, Sven Schreiber wrote:
Ignacio Diaz-Emparanza schrieb:
> El Friday 10 April 2009 01:46:04 Allin Cottrell escribió:
>> I think it's time to "unveil" what we have come up with for
>> user-level access to the Kalman filter, and to ask for bug reports
>> and suggestions for improvement.
>
> Sorry, I have no much time now for testing. I only checked the
> two examples in the pdf help text. All is working ok here. I
> am very impressed by the speedness of the algorithm.
I read the documentation and it looks great, very
straightforward and explicit. Right now I don't have any test
cases at hand and so I will do the same as always --
testing-by-doing when the time comes, and report the bugs.
Thanks, Ignacio and Sven, for the encouraging remarks on the
Kalman filter. I'm currently working on trying to make the
results "Yes" for all entries in the following support matrix
(kfilter = forward pass, filtering; ksmooth = backward pass,
smoothing; ksimul = simulation... and monospace font, please,
maestro!):
Base case Time-varying cross-correlated
coefficients disturbances
kfilter Yes Yes Yes
ksmooth Yes No Yes
ksimul Yes No ?
"Base case" means that the coefficient matrices are time-invariant
and the disturbances in the state and observation equations are
mutually independent. "?" means I think it should work but it's
entirely untested.
Allin.