miken wrote:
> Is anyone aware of a VI that will come up with a set of coefficients
> for a 2-Variable set of data (Bi-Variable Curve Fit, Surface Fit, 2D
> Curve Fit, 2D Least Square Fit with Gaussian Elimination)?
You want to fit an analytic function? That is a regression problem, I
think. Possibly it works by projecting the point set into an
n-dimensional space and then taking the norm of the difference with
respect to an n-plane, where the projection has been proved to be
unique. This provides a metric for the selection of the surface, which
is then fitted to the origional point set with a certain number of
degrees of freedom.
Alternatively, if you can prove that the normal to the surface never has
a negative Z component (assuming the Z-axis is p
ointing up) then every
plane perpendicular to the Z-axis and slicing the surface contains the
projection of a function. For the ZX and ZY slicing planes these are
functions of one variable and can be fit with regular methods. If the
domain is a rectilinear grid of axb points then you can determine a+b
number of curves that define the surface in a local region. For points
inbetween, simply interpolate, assuming that the surface is 'smooth'.
Also, you might be interested in the following link or a spin on it:
http://mathforum.org/epigone/comp.soft-sys.matlab/spoitayzom
