[R-pkgs] New version of systemfit (not backward compatible)

Arne Henningsen ahenningsen at email.uni-kiel.de
Tue Dec 18 17:15:57 CET 2007


Dear R users,

the systemfit package contains functions for fitting systems of simultaneous 
equations by various estimation methods (e.g. OLS, SUR, 2SLS, 3SLS). 
Currently version 0.8 of systemfit is available on CRAN. However, shortly we 
will upload version 1.0, which is NOT BACKWARD COMPATIBLE. The changes that 
broke backward compatibility were necessary to make systemfit() more similar 
to standard regression tools in R such as lm(). We hope that the usage of 
systemfit() is more intuitive for R users now. We will continue to maintain 
the 0.8 branch so that users can still use the old version if they do not 
want to update their R scripts. Both versions are and will be available for 
download from systemfit's website:
   http://www.systemfit.org/
which is a shortcut to 
   http://www.uni-kiel.de/agrarpol/ahenningsen/systemfit/

A paper that describes the (new version of the) systemfit package is 
forthcoming in the Journal of Statistical Software (JSS). A preprint of this 
paper is available on systemfit's website:
   http://www.systemfit.org/systemfit_paper_1.0.pdf

The following list summarizes the most important changes 
from version 0.8 to 1.0:
- some names of systemfit()'s arguments have changed to make it more 
  similar to standard regression tools in R
- the order of systemfit()'s arguments has changed to make it more 
  similar to standard regression tools in R
- the names of the elements in the object returned by systemfit() have 
  changed to make it more similar to lm()
- added several methods for systemfit objects that are generally 
  available for standard regression tools in R
- restrictions on the coefficients can be specified symbolically now
- the functionality of systemfitClassic() has been integrated into systemfit()
- replaced ftest.systemfit() and waldtest.systemfit() by the method
  linear.hypothesis()
- systemfit now uses the "Matrix" package for matrix calculations (this 
  makes the estimation of large models and large data sets much faster)
- improved checking of the arguments so that error messages are more 
  helpful now

We thank two anonymous referees of the JSS, Achim Zeileis, John Fox, William 
H. Greene, Ott Toomet, Duncan Murdoch, Martin Maechler, Duglas Bates and 
several (other) systemfit users for their answers, comments, and/or 
suggestions that helped us to improve the systemfit package.

Feedback is always welcome!
Arne & Jeff

-- 
Arne Henningsen
Department of Agricultural Economics
University of Kiel
Olshausenstr. 40
D-24098 Kiel (Germany)
Tel: +49-431-880 4445 or +49-4349-914871
Fax: +49-431-880 1397
ahenningsen at agric-econ.uni-kiel.de
http://www.uni-kiel.de/agrarpol/ahenningsen/




More information about the R-packages mailing list