non-linear fit

[computing_marmot]

I use the Non-Linear Fit function and, as foreseen in the function description, I enter an initial guess of the coefficients I expect in return.

Then I'm faced to several problems: according to the value of the initial guess entered, the best given coefficients returned by the function vary a lot. As I don't really know which could be a good initial guess, it is a real problem.

Moreover, when I change the initial guess with the "best value" returned once by the program, the value doesn't stand constant, but decreases.

I have tried to enter a large range of initial guesses, and I've noted that when the initial guess is too far from the possible result, the function returns the same values as entered in the initial guess. It seems even not to calculate the coefficients, and just returning the first guess values. I don't understand why it happens, because in the definition of the function, it is specified that if the system doesn't find a solution a the end of the maximum iteration number, it returns an exception error.

Could you help me with these problems, please?

Thanks in advance.

[JorisV]

Hi there,

I have just tried this in LabVIEW with the nonlinear Lev-Mar fit function and I do get an error when the initial guess is too far off. ( error -20041). Also if I use another nonlinear fit function after the first one and use the best guess coefficients as the Initial guess input of the second one than I still get the same coefficients back like in the previous fit. So, in LabVIEW the algorithm works fine and I cannot reproduce the issues you seeing.

Can you send me an example that reproduces the issue in Measurement Studio (That's the soft you are using, right)?

Regards.

Joris V