setting tolerance in lsqcurvefit

Can someone give me an example of setting the tolerance in lsqcurve fit? I am getting a terrible fit and it is quiting very quickly. I think it should be possibel to do through the optimoptions function, but I can find no examples with numerical input.

Thanks in advance

Answer 1

[Star Strider]

Bad fits with the nonlinear parameter estimation routines are most likely the result of either the model not correctly describing the process that created the data, or inappropriate initial parameter estimates.

You have to determine that you defined and wrote the objective function correctly. The Global Optimization Toolbox functions in the Global or Multiple Starting Point Search (link) functions can help with the initial parameter estimate problem.

[Kerry Hipps]

Thank you for your answer. I am sure that the function is appropriate and intial values are close to the desired ones because I call the function with the initial values and plot it over the data defore initiating the fitting.

[Star Strider]

My pleasure.

So it’s not a ‘terrible fit’? Your initial parameter estimates may be the best that lsqcurvefit can estimate.

Answer 2

[Kerry Hipps]

Yes, the actual fit is really terrible with kind of an average line moveing through a sigmoidal like data set.

[Star Strider]

I might be able to help with data, code, and other relevant details.