How to parse different dimension ydata in lsqcurvefit using cell arrays?

Hi. I have compiled all my observed data (of different dimensions) in separate cells e.g.

y{1}=[];
for input_trace=1:h;  % h= no of files from which ydata is imported where filenames are time(1) etc..
    import=['time(' int2str(input_trace) ')'];
    ydata=xlsread(import);  %import of files
    y(input_trace)={ydata(:,2)}; %  
end

and so I have collection of ydata and also my functions output is arranged accordingly.

ini_param=[A,B,Z];  %initial parameters

options = optimset('DiffMinChange',[0.000001],'disp','iter','Algorithm','levenberg-marquardt','MaxIter',100000,'MaxFunEvals',[10000]);%simple max eval fun 100000
options = optimset(options, 'TolX', 1e-14);
[fitted_param,resnorm,fval,output]= lsqcurvefit( @(ini_param,x) functionname(ini_param,x),ini_param,x,y,lb,ub,options);

In this way my ydata is like

y = 

    Columns 1 through 6

      [0.0156]    [320x1 double]    [-0.0212]    [203x1 double]    [199x1 double]    [1.0034e-04]

    Columns 7 through 11

      [246x1 double]    [0.0348]    [242x1 double]    [225x1 double]    [340x1 double]

When I run the program it gives error

Error using lsqcurvefit (line 182)
LSQCURVEFIT requires the following inputs to be of data type double: 'YDATA'.

Error in functionname(line 204)
[fitted_param,resnorm,fval,output]= lsqcurvefit( @(ini_param,x)
functionname(ini_param,x),ini_param,x,y,lb,ub,options);

What am I doing wrong?

Answer 1

[Walter Roberson]

lsqcurvefit does not accept cell arrays for x or y.

You may need to loop doing one y item at a time.

[Vipultomar]

Okay. Thank you. I will give it a try. But where exactly should I use loop. Do I give one cell content (ydata) for one xvalue?

For each xdata values (x1,x2,x3...) I have a number of (y1,y2,y3.. etc) outputs, where each yi is again a vector (but may be of different dimension as I mentioned earlier). So now if I give y's using loop, does that mean it should be like:

[fitted_param,resnorm,fval,output]= lsqcurvefit( @(ini_param,xi) functionname(ini_param,xi),ini_param,xi,yi,l[],[],options);

So that at each x point I do lsqcurvefit. But how do I provide all of my corresponding ydata (for all cells) simultaneously so that it can optimize accordingly.

[Walter Roberson]

Fitting requires that the size of the input (independent variable, each xi) matches the size of the output (each yi). You cannot, for example, use a 57 x 1 vector of x to calculate a 34 x 1 y.

It could make sense to use the same x with different y that are the same size as each other, and it could make sense to use different x with y of corresponding size. For example,

for K = 1 : length(y)
  [fitted_param{K}, resnorm{K}, fval{K}, output{K}]= lsqcurvefit( @(ini_param,xi) functionname(ini_param, x{K}), ini_param, x{K} , y{K}, [], [], options);
end