vector valued guess parameter in fitting function

I would like to fit a function of the form (in myfunc):

ydata = x(1) + ((1 - x(2) - x(3) * log(xdata/f))/v)

to a set of data (using lsqcurvefit or lsqnonlin). the input: ydata is a vector, xdata a matrix. I want to get an estimate of parameter x(2) and x(3) which are constants. x(1) is a constant for every entry of xdata and ydata, meaning it has the same dimension as xdata and ydata, but at some entries I know it is zero beforehand. Does one of the matlab functions allow x(1) to vary with different datapoints? And if so, how do I have to code it in myfunc? All the examples I found so far do not match my problem and I don't know how to call lsqnonlin correctly. Thank you in advance for your help.

sandra

[Matt J]

When you say "constant" it sounds like you really mean "scalar".

[sandrala]

yes, I meant scalar.

Answer 1

[Matt J]

LSQCURVEFIT looks more suitable and that's what I do here, but it should be pretty obvious how to modify it if you insist on using LSQNONLIN

x = lsqcurvefit(@(p,xdata)datamatch(p,xdata,non0Indices,f,v),...
                  p0,xdata,ydata);

function ydata=datamatch(p,xdata,non0Indices,f,v)

   x2=p(1);
   x3=p(2);
   x1=p(3:end);

   X1=zeros(size(xdata));
   X1(non0Indices) =  x1;

   ydata = X1 + ((1 - x2 - x3 * log(xdata./f))./v);