How can I obtain error in parameters estimated in optimization toolboox (lsqcurvefit) ?

I am trying to obtain the value of a parameter by fitting a model equation by lsqcurvefit solver. I give the option show diagnostics and it shows the first order optimality and other values after running the solver. However, as with any optimization problem, I am getting a range of values satisfying the condition. Is there any option through which I can get the standard deviation and mean in the value of the parameter estimated?

Answer 1

[Shashank Prasanna]

It appears that you are looking for confidence intervals to evaluate how good your fit is. Unfortunately LSQCURVEFIT or other functions in the Optimization Toolbox don't have support for fit statistics. However, if you do have Statistics Toolbox you can use the NLINFIT to perform the same fit by rephrasing your model such that you provide the objective function instead of the curve. Here is an example that demonstrate this. http://www.mathworks.com/help/stats/nlinfit.html#btk7kzf-2

Once you have the fit you can use nlparci for the regression parameter confidence intervals for your confidence level: http://www.mathworks.com/help/stats/nlparci.html

hth

Answer 2

[Ketki Sharma]

Thanks. I will try this approach.