LSQCURVEFIT Global Fit Unequal Data Question

Ok, so I am trying to do a global fit of multiple traces to obtain 4 parameters. I have 15 traces that I am trying to fit at the same time using LSQCURVEFIT. Everything works fine if the data in my matrix has an equal number of columns in each of the rows. But, here is the problem. For 6 of the 15 traces, I collected more data. Now, yes I could just delete some of that data so that the number of columns is equal in each row, but my professor wanted more data for the global fit for those 6 traces intentionally. I won't go into detail as to why more data was required for those 6 traces. My question though is how can I do this fit using LSQCURVEFIT if the number of columns for these 6 traces is greater in those 6 rows than the other 9 rows? I'll get errors like index exceeds matrix dimensions (even if the other columns from the other 9 rows is filled with NaN). How can I do this fit with unequal columns in some of my data rows? I read that using cell arrays might work, but I couldn't figure out how with this kind of fit.

[John D'Errico]

You don't need to have the same number of points on each curve to make lsqcurvefit work. But you DO need to show what you did. Otherwise we would need to make a complete guess as to what you are doing, and then write a complete code to solve the problem that we guessed you are trying to solve. That would take FAR more effort than if you just show us what you did!

[Kelly McGuire]

Thanks for the response. Of course, I was planning on showing my work, but not before someone told me it was even possible to do what I described above. Now that someone has, here is my work. I have attached most of the files needed to run my program. Two more files of my data are needed, but they are 11 MB each so I can't attach them here. If you have an email, I can send them to you. As far as the program goes, it is Data.m that is the main program. GlobChiSq is the equation that runs in a for loop and has the 4 parameters that need to be obtained with lsqcurvefit. calcD, calcLambda, and calcPinf are functions called in by GlobChiSq, and make up the rest of the equation. The way the for loop is written (in GlobChiSq), the concentration excel file should remain in column form. The time Excel file needs to be transposed from column format to rows, and time2 can stay as columns (it is just used to plot the experimental data, not used in lsqcurvefit). The other two Excel files that need to be emailed are current, which also needs to be transposed from column format to rows, and current2 which can stay the way it is like time2. It is not used in lsqcurvefit. Notice both the time and current Excel files have one column at the end that has more rows than the rest of the data. This was just a test to figure out if lsqcurvefit could do what I previously described. 5 more traces will be added once we figure out how to do a global fit with multiple traces, some of which have more data points than the others.

[Torsten]

The easiest way to cope with numerical arrays of different length is to use cell arrays:

https://de.mathworks.com/help/matlab/ref/cell.html

Another possible way is to create one long vector of your arrays of different length. The problem in your case is that you also had to save the positions where array (i-1) ends and array i starts because the initial concentrations for these arrays differ.

Best wishes

Torsten.

[Kelly McGuire]

So I tried using cell arrays, but I keep getting the error 'lsqCURVEFIT requires the following inputs to be of data type double'. Is there a way to make the cell array type double?

[Kelly McGuire]

John D'Errico, could you elaborate on your first sentence, "You don't need to have the same number of points on each curve to make lsqcurvefit work."?

[Torsten]

Use "lsqnonlin" instead of "lsqcurvefit".

Then you don't need to use "xdata" and "ydata" as direct input for the solver and you can define them as cell arrays.

Best wishes

Torsten.