Well, I work with complex numbers (their real and imaginary parts) and
non-linear functions. For exemple, the most used theoretical models are
Havriliak-Negami and Cole-Cole (see
http://en.wikipedia.org/wiki/Cole-Cole_equation and
http://en.wikipedia.org/wiki/Havriliak-Negami_relaxation for curiosity).
An important part of my work consist of 'deconvolution' of
experimental data, i.e., decomposing experimental plots as a sum of a
number of these theoretical equations, and sometimes other ones.
This involves non-linear least squares fitting of, say, 9 parameters in
the easiest case. This is easy to do with a 'commercial' spreadsheet, by
using Marquardt-Levenberg or Newton-Raphson algorithms, which use
mathematical tools of matrix algebra (matrix transposes, inverses, etc.)
These routines are available in C and Fortran, but I do not know if they
should be migrated into another language for being used in OOo.
For the sake of code reusability, I need a generic fitting algorithm
regardless of the number of theoretical equations I need to add up, or
their mathematical structure -- I cannot spend my time reprogramming the
equations everytime I need to consider an additional effect.
I do not know if this description is helpful to understand what I need,
but it's just that.
Thank you for all,
Alex
Michael Adams wrote:
On Fri, 08 Feb 2008 12:14:56 +0100
Alex Navarro wrote:
In my daily work I need to fit some complex formulae to experimental
data. This process involves numeric calculus and different fitting
routines, by common least-square minimization algorithms.
Without examples we cannot help much. Do your examples require VBA?
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
