RE: how to fit a graph

[Allen.Windhorn]

Max,

From: address@hidden On Behalf Of Max Rodler

> Can somebody explain me in an easy way how I have to do to fit
> a graph in octave?  I have only a lot of points . That's all. no
> function either.

Fitting a set of data points with a curve based on some arbitrary
function is a bit dangerous, especially if you are going to use
the function to predict something.  Don't you have any idea how
you expect the function to behave?  For instance, if you were
trying to predict power output of a system as a function of power
input, you would expect that the output would be exactly zero if
the input was zero, and the output would never be greater than
the input, and the graph would somewhat resemble a straight line,
but would have added curvature. Then you could select a function
that behaved that way.

Are you trying to fit an output as a function of multiple inputs?
For single-input functions, to explore your data you might try
CurvExpert -- it's a fitting program that tries a large set of
likely functions and gives you the one that it thinks fits best.
<http://www.curveexpert.net>

I also recommend a book called "Fitting Equations to Data," by
Wood and Daniel, and "Numerical Methods that (Usually) Work,"
by Forman S. Acton.

Regards,
Allen