Developing an arbitrary form of equation from parameters and data

Hi, ..., Although Matlab is capable of calculating the equation between data, I am wondering if there is anyway to dictate Matlab to develop our own form of equation like f=-1+(N/Nmax)^q+T/Tmax)^r? Given that N, Nmax, T, and Tmax are known and they are not a single point but a series of data, how q and r can be calculated for the best fit to the equation?

Thanks, Cyrus

[dpb]

Look for the section "Curve Fitting via Optimization" in the documentation under Matlab/Mathematics/Optimization. It gives an example.

If you have the Optimization and/or Curve Fitting toolboxen there are additional tools therein.

[Walter Roberson]

Cyrus commented "Thanks mate, your comment was so helpful. Cyrus."

Answer 1

[Star Strider]

If N, Nmax, and T are vectors, you first need to use meshgrid to convert them to matrices with the appropriate orientations. Then you can use nlinfit or lsqcurvefit to do the regression. There is an example here for fitting a surface. You can expand it to your problem.

You can also experiment with the more general Statistical Toolbox functions listed under Nonlinear Models.

[Kourosh]

Thanks mate for your great help.

Cyrus

[Star Strider]

My pleasure!