Scientists often have to correlate analytical data to complex
parameterised functions. There are a number of well known techniques
which fit a given function to a set of data points (lookup Marquart in a
Numerical Recipes book). Given a set of data, a function and an
estimate of the functions parameters, it is possible to reduce the
root mean square error in an iterative fashion.
I originally wrote a curve-fit program for an Apple II in Pascal. This
was ported to MS Pascal for an IBM PC running DOS and rewritten under
Turbo Pascal, also for DOS. A friend of mine ported this code to C
for Windows 3.1. It turns out that the user interface is far the most
challenging aspect. The calculation is relatively straight forward and
I would be very interested in a Java version of this program for
teaching purposes (in other subjects). It would also be a nice
demonstration piece - far too many Java demos lack substance.
I'm willing to supervise one student for this project.