Nonlinear least squares fit of custom function to data

Dear all,

I'm working on describing a physics model using the following function: [y(i)] = f ([x(i)], c1, c2, c3, c4, c5, c6, P) where P is a struct that contains even more parameters.The function is written in its own .m file and contains quite a body of calculations. What I want to do is a nonlinear least squares fit on a given dataset x-y using the function f and determine c1, c2, c3, c4, c5, c6. The struct P contains known parameters. I can't seem to get my head around passing f to fittype or lsqcurvefit for that matter.

My 1st approach:

fitting = @(c1, c2, c3, c4, c5, c6,x) f(x,c1, c2, c3, c4, c5, c6,P); ft = fittype(fitting, 'independent', {'x'}, 'dependent', {'y'}); opts = fitoptions( ft ); opts.Display = 'Off'; opts.Lower = [-Inf -Inf -Inf -Inf -Inf -Inf]; % lower bound of coefficients to be fitted opts.StartPoint = [0 1 2 2 (10^-4) 1]; % starting points of coefficients opts.Upper = [Inf Inf Inf Inf Inf Inf]; % upper bound of coefficients to be fitted % Fit model to data. [fitresult, gof] = fit( [x, y], ft, opts );

This code fails on the fittype call producing the error: 'Custom equations must produce an output vector, matrix, or array that is the same size and shape as the input data.This custom equation fails to meet that requirement' I don't see why is this the case since f was defined to produce a vector y for a given vector x and some parameters.

 My 2nd approach:

Define f as: f'([x(i)], p, P) replacing c1, c2, c3, c4, c5, c6 in the function's body as elements of vector p: p(1), p(2), p(3), p(4),p(5), p(6). Then called:

[p,error] = lsqcurvefit(@f',[0 1 2 2 1 1],x) but the execution of f' fails with error 'not enough input parameters'. It seems the struct P is not passed in f'

This post got long as I tried to be slightly thorough, but hopefully someone can give a pointer or two to help me get unstuck.

Thanks in advance!