Skip to main content
payman1-Visitor
1-Visitor
Dec 3, 2014 12:30 AM UTC
Question

genfit - improve the results

Hello All,

I am using "genfit" routine to fit my data to a modeled signal. The modeled signal in the attached worksheet is sp(k,b) where "k" is the variable and "b" is a set of coefficients.

My first question is that : the output of the "genfit" does not perfectly match with the original data. Is there anything I could do to improve the results? (Except changing the guess values, because they are obtained from another section of the program)

For the second question, the residual function highly depends on the variable "k" as is shown by the graph. Is it a normal behaviour?

Thank you very much in advance

6 replies

RichardJ19-Tanzanite
19-Tanzanite
Dec 3, 2014 5:19 PM UTC

If you zoom in on the graph of Y and sp(k,bb) you will see that at the start Y leads sp(k,bb), in the middle there is almost no phase difference, and at the end sp(k,bb) leads Y. So you have a gradually varying phase, which is not allowed for in your fit. That's why the residuals look the way they do. Since your "data" is synthetic you will have to figure out why it is not modeled well enough by your function. I suspect that it has something to do with the fact that your data is created using 6 parameters, but you are only using 5 parameters in the fit (b0 is not used anywhere in your function).

Translate
payman1-VisitorAuthor
1-Visitor
Dec 3, 2014 5:23 PM UTC

Thank you Richard, b0 is constant everywhere. I replaced that with n0=1.337 during "genfit" process.

Translate
RichardJ19-Tanzanite
19-Tanzanite
Dec 3, 2014 5:28 PM UTC

Your guess values are too far off. If you make b:=inn for the guesses, the fit is much better. Genfit is finding a local minimum.

Translate
Terms & ConditionsAccessibility statement