Meaning of W_sigma when performig a fit with constraints
martina
Tue, 10/18/2011 - 07:55 am
I have a problem in understanding how the W_sigma in the fitting functions is calculated. I have to fit a data set with two gaussians and I create therefore with FuncFit an appropriate function to fit them which is the sum of the two gaussians.
The only constrain which I want to impose is that the X02 (central position of the second gaussian) has a fixed distance with respect to the X01 (central position of the firts gaussian) and let's say the distance is 3.
I do this by using a constrain like this X01+3 smaller then X02 smaller then X01+3. The fitting seems working if I look at the X01 and X02 results, but I cannot understand the W-sigma values associated with this variables. I would expect them to be the same because X02 is not independent from X01, but this is not the case. What is the algoritm doing if i choose such a constrain? X02 in my intention should at each fitting step be at a fixed distance with x01, but since I get out an error on it this suggests that is varying indipendently. Is that possible?
Martina
A better approach to what you want to do would be to re-write your fitting function to use x0 as the location coefficient for the first Gaussian, and (x0+3) as the location of the second Gaussian. If the offset needs to be settable, you can make the offset a fit coefficient and then hold the value of the offset.
John Weeks
WaveMetrics, Inc.
support@wavemetrics.com
October 18, 2011 at 11:30 am - Permalink