Support for Box-Jenkins model structure

[Hofsmo]

Hi,
I was wondering if there are any plans on adding support for the Box-Jenkins model structure? For some of my previous research I used the implemenentation from the System Identification Toolbox in MATLAB quite a lot, but I never found a good open source alternative.

[baggepinnen]

Hi!

Not really, a black-box statespace model structure can represent any Box-Jenkins structured model, so I haven't seen any good-enough motivation to adding additional support for those Box-Jenkins and instead recommend using newpem to estimate a general statespace model. With a statespace model from newpem, you can decompose this model into the part driven by input and the part driven by noise using innovation_form, the first example here demonstrates that Validation.

Having said that, the structured_pem method is able to estimate any structured linear model, but to use it you'd have to take the parameter vector and turn that into your BJ model and finally convert that into statespace form using ss in the user-provided constructor function.

[Hofsmo]

Many thanks for the quick reply. I don't have any experience with state space identification, but I will check it out :) .

[baggepinnen]

Please be encouraged to reach out if any questions appear!