Global fit does not compute on all my data range

Share icon

Hi,

I am trying to make a global fit over two different datasets :

•print WE

WE[0]= {120.61,144.92,191.86,209.37,216.11}
•print ME
  ME[0]= {47.49,76.78,127.72,151.19,205.11}

versus:

•print Conc
  Conc[0]= {100,250,500,750,1000}

with the following function:

return w[0]-w[1]*exp(-(x-w[2])/w[3])

Here is the output of the Global fit (A=w[1] linked) that I got in the Command window:

  *** Doing Global fit ***
  Fit converged properly
  Duplicate/O YCumData,WMCF_TempXWave
  WMCF_TempXWave = x
  NewGlblFitFunc(MasterCoefs,FitY,WMCF_TempXWave)
  FitY=FitY[p]
  NewGF_ResidY = YCumData[p] - FitY[p]
  KillWaves/Z WMCF_TempXWave
  MasterCoefs={254.96,125.53,145.22,603.79,993.14,11047,5446.1}
  V_chisq= 76.3488;V_npnts= 9;V_numNaNs= 0;V_numINFs= 0;
  V_startRow= 0;V_endRow= 8;
  W_sigma={12.1,5.36e+05,2.58e+06,90.8,1.94e+03,2.33e+07,1.24e+04}
  Coefficient values ± one standard deviation
  	K0	=254.96 ± 12.1
  	K1	=125.53 ± 5.36e+05
  	K2	=145.22 ± 2.58e+06
  	K3	=603.79 ± 90.8
  	K4	=993.14 ± 1.94e+03
  	K5	=11047 ± 2.33e+07
  	K6	=5446.1 ± 1.24e+04
  Global Fit converged normally.

 
Global fit results
  	Fit converged normally
  V_chisq =  76.3488  V_npnts=  9  V_numNaNs=  0  V_numINFs=  0
  Number of iterations:  2
  Data Set:   root:WE   vs   root:Conc  ; Function:   ExpManual
  	0	yinf[ExpManual][root:WE]	254.956 +- 12.059
  	1	A[ExpManual][root:WE]	125.526 +- 535545
  	2	xo[ExpManual][root:WE]	145.219 +- 2.57603e+06
  	3	tau[ExpManual][root:WE]	603.794 +- 90.8227
  Data Set:   root:ME   vs   root:Conc  ; Function:   ExpManual
  	0	yinf[ExpManual][root:ME]	993.142 +- 1941.43
  	1	A[ExpManual][root:WE]	125.526 +- 535545 ** LINKED to data set root:WE coefficient 1: A[ExpManual][root:WE]
  	2	xo[ExpManual][root:ME]	11047.1 +- 2.32671e+07
  	3	tau[ExpManual][root:ME]	5446.13 +- 12358.8

Each dataset is 5 elements long (so 10 total), but somehow the Global fit only computes over 9 elements. Would you see any explanation of why this would happen?
Thank you!

It's possible that you posted enough information to suggest something, but I'm not sure about that. Could you post your Igor experiment file? If you can do that, then I might be able to take a look at exactly what you did and figure out what's going on.

May 28, 2020 at 10:30 am - Permalink

As you sent the experiment file, the weighting panel has two non-existent waves selected as weighting waves: WTEarea_err and MTEarea_err. There are four waves with the suffix "_err"; two of those have "perim" as part of the name, so I expect those are the ones you are using. They each have a NaN (not-a-number, in an Igor table, a blank cell) at the end. If you are using those weighting waves, the NaNs will mask away the last of the five points from each wave when you do the fit.

June 3, 2020 at 12:07 pm - Permalink