The official release version is 2.2b, which is compatible with Ox 3.x and 4.x. Additionally, it fixes some minor issues in the example programs. In order to use SsfPack, you need to have a working installation of Ox 3.0 or higher (see Jurgen A Doornik work page). To install SsfPack you need to download the file ssfpackV22b.zip (918 KB) and unzip it from the main Ox directory (for Windows, this is usually c:\Program Files\Ox\). The Windows version can be directly downloaded from this link.
Unix versions of SsfPack are available from http://www.doornik.com. Note that these only include shared libraries. In order to use them, the header file ssfpack.h (and optionally the documentation and examples) from the Windows version must also be installed.
It is important to stress that SsfPack 2.2b, like the console version of Ox, may be used freely for academic research and teaching purposes only. Commercial users and others who do not qualify for the free version must obtain a license from S.J. Koopman.
After extracting the package in the Ox directory, the following files will be present:
\packages\ssfpack\SsfPack.dll (the SsfPack 2.2 library) \packages\ssfpack\SsfPack.h (the SsfPack 2.2 header file) \packages\ssfpack\SsfCombine.ox (ox code combine functions) \packages\ssfpack\code\airline.mat (data file) \packages\ssfpack\code\nile.mat (data file) \packages\ssfpack\code\spirit.bn7 (GiveWin data file) \packages\ssfpack\code\spirit.in7 (GiveWin data file) \packages\ssfpack\code\*.ox (ox code examples) \packages\ssfpack\doc\SsfPackV22.ps (documentation PS format, 1053KB) \packages\ssfpack\doc\SsfPackV22.pdf (documentation PDF format, 508KB)
You should now be able to use SsfPack. You may wish to try the example Ox program Ssfkf.ox which should give the output
Ox version 3.40 (Windows) (C) J.A. Doornik, 1994-2004
mKF' (t=10)
v K(1,1) K(2,1) F^-1
-3.0347 0.76491 0.21161 0.44669
Have fun with SsfPack 2.2 !
SsfPack 3 contains many new functions and improvements over version 2.2. For example, it uses exact initialisation algorithms for state space models with diffuse elements in the state vector, rather than the numerical approximations that are used in version 2.2. It also handles the system matrices as sparse matrices. Further, a new form of the filtering and smoothing equations greatly improves the computation speed for high dimensional multivariate models. These developments have led to a much more stable and faster version of SsfPack. Moreover new functions are provided to put (seasonal) ARIMA models and generalised unobserved component time series models in state space form, to compute the multi-step likelihood, to do forecasting and to simulate from smooth densities.
SsfPack 3 is a commercial product and can only be obtained from Timberlake Consultants Ltd or from S.J. Koopman.