uncertainty and curve fitting

Hi,

Please I need your help. I'm working on curve fitting, I'm using lsqcurvefit function to do it!! I'm trying to estimate uncertainty of the coefficient A and B of the function fitted to my observation ponits (y=A.x^B)!! Please could you help me!! Thank you in advance

Mira

Answer 1

[Richard Willey]

If you need confidence intervals nlinfit is a better option

Answer 2

[bym]

this example shows how to bootstrap to get the standard error in the coefficients. You can adapt it to use lsqcurvefit or transform your model to linear using logarithms

load hald
x = [ones(size(heat)),ingredients];
y = heat;
b = regress(y,x);
yfit = x*b;
resid = y - yfit;
se = std(bootstrp(...
         1000,@(bootr)regress(yfit+bootr,x),resid));

[Richard Willey]

Bootstraps are great. I love them to death. However, I question whether they're an appropriate solution when parametric methods are available.

In general, I think of bootstraps as something we do out of necessity when a parametric estimate isn't feasible. For example, generating confidence bounds around the median, bootstrapping a LOESS curve or a kernel smoother.

Answer 3

[Mira]

Thank you guys for your help. This is exactly what I'm looking for!!! Thanks