MATLAB: How to apply a model to a dataset and calculate coefficients using non-linear regression

non-linear regression

The problem is attached, and I'm just not sure how to calculate the coefficients. Do I use polyfit?

 Can someone show me what to do?

Best Answer

You could possibly use polyfit by setting ‘x=1/m’ and regressing against exp(G), then re-transforming it later. Such transformations are generally to be avoided, because they also transform the errors, and lead to inaccurate parameter estimates.

However, I get the impression that the assignment wants you to use fminsearch, nlinfit, or lsqcurvefit. You can easily code the regression as an anonymous function to use with these optimisation functions:

G_fcn = @(B,m) log(B(1)./(m.^2) + B(2)./m + B(3));

I tested this with nlinfit and your data and got a good fit. Since this is your assignment, I leave the rest to you.

Related Solutions

MATLAB: Error while opening a simulink model

Just Change the extension in mdl instead of MDL.

MATLAB: How to normalize data from a periodogram

I solved the problem using linear regression:

The points below the blue line receives zero, and the value points above the blue line are subtracted by the value of the same point in the linear regression. Something like this:

if data[x] < linearRegression[x]
    data[x] = 0
else
   data[x] = data[x] - linearRegression[x]
end

Note: in this case I don't care about the value itself, I just need to see where the peak occurs.

Best Answer

Related Solutions

MATLAB: Error while opening a simulink model

MATLAB: How to normalize data from a periodogram

Related Question