Surface Fitting Tool Coefficient Structure

Hello,

I am trying to fit a surface for a group of data. I defined input matrices x and y; and output matrix z on the command window

x = [50;75;90;50;75;90;50;75;90];
y = [80;80;80;114;114;114;160;160;160];
z = [18.76;23.67;19.19;35.81;48.74;32.74;22.17;36.21;37.26];

and I called sftool. In polynomial surface fit, I can see the function f(x,y) and its coefficients. However, I cannot see the function and its coefficients when I select Interpolant or Lowess.

When I select Interpolant tool with cubic method, I cannot find the coefficient of function.

Is it possible to see the function fof fitted surface?

I am using R2009b version

Thank you,

Fevzi YILMAZ

Answer 1

[Richard Willey]

Local regression (aka LOWESS/LOESS) and interpolation don't lend themselves to parametric representations. There really isn't any simple set of coefficients that you can use to describe your model.

In general, when I see people asking for an equation or a set of coefficients that describes the model its because the need to share the results with a third party.

The easiest way to share your non parameteric fit to users without MATLAB is simply provide them with the object that you've created. If your customer needs to understand the dynamics of the system, they can use the plot method and inspect the resulting fit.

If you need to share your results with a non-MATLAB user, I normally ask whether they're able to work with some kind of look-up table. If they can, its simply a matter of choosing the right number / location for a set of break points.

[Fevzi]

Thank you for your answer.