Welcome State space notation Functions Download
Last update Jul 10, 2004

This library was developed while I was writing my Ph.D. Thesis titled "State Space Modeling of Exponential Family Data". The library is available for both S-Plus and R. I recommend R as it executes the scripts approximately 25 times faster than S-Plus (and it is free ;-)

The main feature of the library is the implementation of the iterated extended Kalman filter and smoother to discrete data from an exponential family. The latent process is assumed to be linear and Gaussian.
The following families and link functions are supported:
Binomial       identity, logit, probit
Gaussian       identity
Poisson        identity, log
Multinomial       canonical, pom

State space models are a very general class of models covering a range of standard models, like random effects models, ARIMA models, structural time series equations and so on. However, this very generality of the state space models makes it difficult to implement a general method for handling every possible situation.
I have therefore, as far as possible, used the notation from my thesis in all implemented scripts. My notation is highly inspired by West & Harrison (1997). It is therefore my hope, that it should be reasonably simple to modify the scripts to handle special cases, for instance to handle non-linear state equations.
I have scrutinized my code, but the usual disclaimer applies -
bug-reports are always welcome.
Bjarke Klein, May 2003.