The Academic and NAG library editions

The previous versions 1, 2, 3, and 4 of Simfit used the NAG library extensively for numerical analysis, but versions 5 onwards have not until recently been able to call the NAG library for number crunching. That situation has now changed with the availability of both academic and NAG library editions. The current editions of Simfit use some 215 NAG library numerical analysis subroutines, and the two versions of Simfit differ mainly in the way these are implemented.

• The Academic edition
  In this version, the numerical analysis codes in the dynamic link library w_maths.dll (and x64_maths.dll) have been constructed using tried and tested public domain code where possible, edited to correspond to the NAG library subroutine arguments. Otherwise code written by W.G.Bardsley has been used.
  This version is currently supported and being developed.
   
• The NAG library edition
  This version uses a link between the dynamic link library w_maths.dll and the NAG dynamic link libraries.
  It can only be used if you have working versions of the NAG DLLs installed on your PC.
  However, note that if you do not have the NAG DLLs installed on your PC, you can download a trial version from the NAG library website.
   
• Availablity of alternative versions
  The academic and NAG library editions are both free, but the NAG library edition provides additional features.
   
• Advantages of the NAG library edition
  Not only does the NAG library version use NAG software for numerical analysis, but it has also been enlarged in the areas of optimisation, differential equation solving, curve fitting, and documentation.
    For instance, extra procedures in the NAG edition include the following.
   
  1. Faster implementation of BLAS and LAPACK.
  2. Regression on ranks is provided with options for strata and right censoring.
  3. Bounded influence regression is available to calculate M-estimates by all the usual robust regression techniques.
  4. The NAG optimisation (E04) and differential equation solving (D02) routines can be used.
  5. IFAIL messages are written to the file nagifail.txt, which can be browsed retrospectively in the case of failure of iterative algorithms to converge and similar problems.
  6. There is a mechanism to demonstrate how to use the NAG library functions and data sets from within the Simfit environment.

Additional documentation

Read nag_simfit.pdf for a description of how to use the NAG library for data analysis.