# MultiPeakFit with constraints wave

ukos

I am using igor's multi peak fit (Do_MPFit()) to fit a 2d spectrum automatically. Therefor I extract every single line and hand it in a for loop to the Do_MPFit() Function but I am stuck right now. I think it all boils down to two problems, where I can't find help in the Igor Help Files:

I am using Lorentian or Voigt fit functions. In the Analysis->CurveFitting Dialog there are 4 initial parameters required for lor-Functions. In the MultiPeak Procedure Panel there are only 3 required.

How many Parameters do I need for the Do_MPFit() function to work with Lorentian/Voigt/Gaussian?

I'd like to set constraints for the peak positions but when I give my constraints wave to the Structure used by Do_MPFit(), it stops to do anything and the parameters sty the same. Probably because the constraints are set to the wrong parameters. It would help very much to see the Lorentian or Voigt Function that is beeing used by Igor (MPFXLorenzianPeak) but I can't find it.

I created a minimum working example of my code. There are two peaks beeing created and a multipeak fit should reproduce this function. As I understand ist, there are two peak-functions with increasing K values and a linear background:

* y = K0+K1x

* y = K2+K3/((x-K4)^2+K5)

* y = K6+K7/((x-K8)^2+K9)

Now I can set the constraints for them as K4>950 and K4<1000. Somehow this doesn't seem to work properly. My MWE-Code is attached below.

```
//minimal working example for use of constraints
#pragma rtGlobals=3
#include <Multi-peak fitting 2.0>
Menu "Testing" //create menu bar entry
"MWE",MPPM()
End
Function MPPM()
String gstrDataFolderCurrent = "root:PLEM"
Wave/Z peak65, peak76, peakBG, wavToFit
Struct MPFitInfoStruct MPPMfit
//Make new working dir
NewDataFolder/O/S $gstrDataFolderCurrent
//Instantidate peak waves
make/O/N=4 peak65 = {0,10,950,100}
make/O/N=4 peak76 = {0,10,1090,100}
make/O/N=2 peakBG = {0,0}
// y = K0+K1x
// y = K2+K3/((x-K4)^2+K5)
// y = K6+K7/((x-K8)^2+K9)
make/O/T/N=6 constraints = {"K0 > 0", "K1 > 0", "K2 > 0", "K4 > 936.12", "K4 < 986.12", "K8 > 1082", "K8 < 1132"}
make/O/N=1000 wavToFit
SetScale/I x, 900,1300, wavToFit
wave wavToFit
wavToFit = 1/pi*x*(10)/((x-961.118)^2+(10)^2) + 1/pi*x*(10)/((x-1107.0)^2+(10)^2)
//INIT MultipeakFit
wave MPPMfit.yWave = wavToFit
MPPMfit.NPeaks = 2
MPPMfit.FitCurvePoints = 200
MPPMfit.fitOptions = 4
MPPMfit.ListOfFunctions = "linear;Lorentzian;Lorentzian;"
MPPMfit.ListOfCWaveNames = "peakBG;peak65;peak76;"
MPPMfit.ListOfHoldStrings = ";;;"
// wave/T MPPMfit.constraints = constraints
if (MPF2_DoMPFit(MPPMfit, (gstrDataFolderCurrent+":")))
print "Abort due to Error: " + num2str(MPPMfit.fitError)
return -1
endif
if (MPPMfit.fitQuitReason)
print "Abort due to fitQuitReason: " + num2str(MPPMfit.fitQuitReason)
endif
print "FuncListString: " + MPPMfit.FuncListString
print "chi: " + num2str(MPPMfit.chisq)
print "Number of Points used in fit:" + num2str(MPPMfit.fitPnts)
End
```

I'm not familiar with MPPMfit but your coefficients K0, K2, and K6 are linear dependent. Together with the information about the 3 or 4 required parameters my guess would be that Lorentzian takes no offset (is included in "linear").

HJ

January 23, 2015 at 06:38 am - Permalink

* y1 = K0+K1*x

* y2 = K2/((x-K3)^2+K4)

* y3 = K5/((x-K6)^2+K7)

* y = y1 + y2 + y3

I could also live with defining my own Lor/Gaus/Voigt Fit Function but also don't yet know how to do it inside the MultiPeakFit Environment. Maybe I have to drop the MultiPeakFit Environment and go back to the original FuncFit. At least FuncFit seems to work a bit more transparent.

January 23, 2015 at 08:41 am - Permalink

And yes, if each peak function defines its own vertical offset coefficient, they are linearly dependent. Consequently, the peak functions for MPFit2 do not have that coefficient, leaving that instead to the various baseline fit functions.

Setting up the constraint wave for a multipeak fit is pretty difficult. I would recommend setting up one fit in the MPF2 GUI and examining the resulting wave.

John Weeks

WaveMetrics, Inc.

support@wavemetrics.com

January 23, 2015 at 09:28 am - Permalink