What tool to use to perform Nonlinear Regression with more than 3 variables?

Hi

I am trying to perform a nonlinear regression in Matlab. I have one dependent variale (response) and 16 independent variables (predictors). I am trying to find any tool in Matlab that can perform the nonlinear regression or curve fitting for all of them together.

Is there any such tool in Matlab?

[Sebastian Hahn]

fitnlm - Function

Answer 1

[Star Strider]

The nlinfit function will do what you want.

The procedure for having multiple independent variables is not quite obvious, but easy enough to do. The nonlinear fitting functions will only take one argument for the independent variable, but that argument can be a matrix. To do a nonlinear regression with multiple independent variables, combine your different independent variables into a matrix, and pass that to nlinfit.

For example:

% EQUATION: y = exp(q*x1) * sin(r*x2)
% MAPPING: b(1) = q,  b(2) = r,  x(:,1) = x1,  x(:,2) = x2
x = [x1(:), x2(:)];
f = @(b,x) exp(b(1).*x(:,1)) .* sin(b(2).*x(:,2));
B0 = [1; 1];
B = nlinfit(x, y, f, B0);

(Note: Untested code, for illustration purposes only!)

[Ruby Wei]

What if two indepent variables have different dimansions? In this case, they can't be combined into one matrix.