Hi, let me preface this by saying that I like this package and using it has generally been great.
The PEM section of the state-space identification documentation says the following:
A trade-off between prediction and simulation performance can be achieved by optimizing the h-step prediction error. The default is h=1 which corresponds to the standard prediction error. This can be changed using the keyword argument h. A large value of h will make the optimization problem computationally expensive to solve.
However, the newpem function doesn't actually take h as an argument according to the documentation string.
Either I've not understood this correctly or one of them is wrong.
Related question: is it currently possible to use newpem with multi-step prediction error?
Hi, let me preface this by saying that I like this package and using it has generally been great.
The PEM section of the state-space identification documentation says the following:
However, the newpem function doesn't actually take
h
as an argument according to the documentation string.Either I've not understood this correctly or one of them is wrong.
Related question: is it currently possible to use newpem with multi-step prediction error?