Hi Willy,

I'm not sure what you mean exactly - perhaps you could elaborate? The function lsqnonlin(FUN,X0,...) will internally minimize SUM(FUN(X).^2) and thus convert the function values to a scalar, so the size or shape of FUN(X) is not relevant in itself.

Thank you very much. I agree. I think that the problem is the syntax. I need to do a global fit of data within an array. For example, I have two independent variables, voltage, and drug concentration. Each member of the array is a matrix containing one list of voltages and the open probability at one drug concentration. So, the number of elements of the array is the number of different drug concentrations tested.

I am trying to do a global fit, my anonymous function starting in this way:

fun=@(Parameters)data_array{:}(:,2)-1./(1+...

(because my data are in the second column of each member of the array).

But it not work. It produced the error: Bad cell reference operation.

Thanks for the clarification. The syntax you are using to refer to the second columns of the matrices in the cell array seems to be erroneous. I would recommend writing a simple wrapper function that converts the output to a vector, and then inputing that to lsqnonlin.

Thank you Pekko. I did almost the same: I converted the members of the array into a big matrix. But I am still curious about using lsqnonlin for arrays

That sounds like a sensible solution. I would avoid using cell arrays in numerics if they are not necessary, because they may make the code less efficient than with just numerical arrays, even if one manages to make the functions work with them.

I agree. But I needed to use the array to make independent fit at each different drug concentration (the number of rows are different at each member of the array). Thanks again. Cheers