Fitting matrix error

[Noor1]

Dear All,
I am trying to fit some data point, using a fitting function.
The function I use and the fitting result are here in

TF1 *eff_function= new TF1("eff_function","[0]*exp(-[1]*(log(x-[2]+[3]*exp([4]*x))))",50,1900);
 
    eff_function->SetParameter(0,2);
    eff_function->SetParameter(1,0.80);
    eff_function->SetParameter(2,-134);
    eff_function->SetParameter(3,10313);
    eff_function->SetParameter(4,-0.02);

result

 FCN=4.9035 FROM MIGRAD    STATUS=CONVERGED    1456 CALLS        1457 TOTAL
                     EDM=7.31815e-07    STRATEGY= 1      ERROR MATRIX ACCURATE 
  EXT PARAMETER                                   STEP         FIRST   
  NO.   NAME      VALUE            ERROR          SIZE      DERIVATIVE 
   1  p0           1.17188e+02   4.85222e+01   1.09204e-03  -3.93999e-04
   2  p1           8.07679e-01   5.54103e-02   1.46099e-06   3.08823e-01
   3  p2          -1.00193e+02   3.89770e+01   4.21887e-03  -6.25063e-05
   4  p3           1.05498e+03   1.66706e+03   1.78960e-01  -4.90144e-06
   5  p4          -4.81772e-02   2.79259e-02   2.64344e-06  -2.75212e-01
chi2  4.9035

when I plot the fitting function it looks good and match all the points, but when check the error matrix I found large percentage error in the parameters, as well as the chi2??

does this mean the fitting is not good enough?? or am I consider wrong error parameters??
I have attached the fitting plotfit.pdf (15.6 KB)

[moneta]

Hi,

It could be you are overfitting. It looks like given the large error and not so many data points the function has too many parameters and it follows exactly the data point.
It is therefore expected a very small chi2 and large errors on some of the parameters

Lorenzo

[Noor1]

thank you for reply,
but this is the point of the fitting, I want the function which follow the points I have, I tried different functions, and this is the function that fit.
is there a way to reduce this large errors??

[Wile_E_Coyote]

You could try to use “ME” fitting options.