Y fit error for weighted levenberg marquart algorithm

3 views (last 30 days)
I am using the weighted levenberg marquard algorith, and I want to calculate the error for each fitted x value. I have a vector of x data, a vector of y data, and a vector of standard deviation of the y data. Each vector has n points. For each data point there were several measurements (at the same value of x), so I calculated the stdev of these y values (at each x value). The wights are 1/ stdev_ymeasured^2.
After performing the fit, how can I calculate the error (or the uncertainty of the fit) for each input value of x. The fit is obviously a smooth line, but the uncertainty should vary along it, both because the weights are not equal and because the residuals at each point are different.
I tried using eqn. 19 from this page: http://people.duke.edu/~hpgavin/ce281/lm.pdf but the dimentions of W are wrong; I assume W is a 1*n matrix, with each entry wi being the weight for each x value. Also, I am total unclear as to what sigma_y^2 is supposed to be. Is there a better/different way to calculate the uncertainty? If this formula is appropriate, what is the matrix W actually supposed to be?
  1 Comment
chu li
chu li on 10 Mar 2020
Hello, now i'm tring to implement the weighted levenberg marquard algorithm. But i don't know how to add the weight to the fitting program. Can you show the program of weighted levenberg marquard algorithm? Thank you very much.

Sign in to comment.

Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!