cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
Showing results for 
Search instead for 
Did you mean: 

Community Tip - Learn all about PTC Community Badges. Engage with PTC and see how many you can earn! X

genfit problem about number of parameters

jcoolman
3-Visitor

genfit problem about number of parameters

The error message says the number of fitting parameters and guess values must be the same, but they do !!!

Please, take a look to the file attached, this is a short programm.

1 ACCEPTED SOLUTION

Accepted Solutions

Not exactly sure what you mean by "The combination of these three functions gives a family of non-asymptotic sigmoid functions". Which kind of "combination"?

If you have a simple additive combination in mind, you would have to do it that way, nut the result sure is not satisfactory (and very sensitive wrt the guess values):

Werner_E_0-1606247813739.png

But if you want to say that you would like to split your curve into three parts and fit a different function type to each part, then you would have to do the splitting yourself and fit three times.

 

BTW, Here's a function which seems to fit your data pretty well:

Werner_E_1-1606248764584.png

 

 

 

View solution in original post

6 REPLIES 6

Not exactly sure what you mean by "The combination of these three functions gives a family of non-asymptotic sigmoid functions". Which kind of "combination"?

If you have a simple additive combination in mind, you would have to do it that way, nut the result sure is not satisfactory (and very sensitive wrt the guess values):

Werner_E_0-1606247813739.png

But if you want to say that you would like to split your curve into three parts and fit a different function type to each part, then you would have to do the splitting yourself and fit three times.

 

BTW, Here's a function which seems to fit your data pretty well:

Werner_E_1-1606248764584.png

 

 

 

Yeah, I was talking about "linear combinations of functions" , sorry.

Thanks, you helped me to fix this error, now I can fit with genfit.

I think mathcad is not good for fitting points if they are not well close to a mathematical curve, but experimental points are random up and down from such a curve.

I was hopping to fit experimental points to non asymptotic sigmoid curves like this https://stats.stackexchange.com/questions/398551/logistic-function-with-a-slope-but-no-asymptotes

(because some chemical law predicts this) you see there are 5 paramenters, I also added a sixth parameter x0  in this form: instead x I put (x-x0) hopping the fitting process could find the best x0, but the outcome was ugly.

SigmaPlot can fit this without problem (sorry for comparison) but at the moment I haven't this software.

Thank you anyway.

Mathcad often is very sensible with respect to the guess values which will have to be chosen with great care.

You may want to have a look at the eval version of CurveExpert Pro. Even in demo w/o the various export options mode its quite useful for curve fitting. It has a lot of predefined function types which it tries on the data and you may also define your own.

FredLusk
12-Amethyst
(To:Werner_E)

+1 for CurveFit Pro.  One thing I have been using it for lately is reverse engineering a few older nomographs that are used in my type of civil engineering, but for which I cannot find the equations that were used to build them.  I extract data points from the curves with a scale, drop them into CurveFit Pro, and find the best fit equation that I can then use in Mathcad, Excel, or one of my programmable HP calculators.

Could not find "CurveFit Pro" in a first quick search on the net. Did you mean "CurveExpert Pro" which I mentioned above?

FredLusk
12-Amethyst
(To:Werner_E)

Werner...

 

You are correct.  I brain-freezed the name.

 

Fred

Top Tags