FitMySpectrum

Share icon

I have a equation with two fit coefficients to fit data from a 2D wave. I want to display all fitting curves along with the raw data and store coefficient values in wave form; somewhat similar to batch curve fitting. There are few related forum posts that discusses about automatic curve fitting and how coefficients are stored in data folder however; due to my limited knowledge I am having trouble to follow the coding and hence I am trying to work with my code. Right now, my code does not accept the input wave to which I want to fit my data. Can anyone suggest/help me to make my code work?

#pragma rtGlobals=3     // Use modern global access method and strict wave access.
#include <WMBatchCurveFitIM>

Function FitMySpectrum(ww, xx) : FitFunc

    WAVE ww    
    Variable xx
    Variable w1=0,w2=0,w3=0,w4=0,w5=0,w6=0,w7=0,w8=0,w9=0,w10=0,w11=0,w12=0,w13,w14=0,qq=0
    Variable T1=0.24, S1=0.047, T2=0.24, S2=0.047, x1=1.91, x2=1.918, S3=0.26, x3=2.20
    Variable k, i
    Make/O/N=46 A1   //wave to store coefficients A1
    Make/O/N=46 A2   //wave to store coefficients A2
    Make/O/N=(41,46) fitdata    //wavet to store results from fitting
                             
    //CurveFitDialog/ These comments were created by the Curve Fitting dialog. Altering them will
    //CurveFitDialog/ make the function less convenient to work with in the Curve Fitting dialog.
    //CurveFitDialog/ Equation:
    //CurveFitDialog/ f(xx) =  (A1*(1110*(exp(0.5*(S1/T1)^2-(xx-x1)/T1))*erfc(0.707*(S1/T1-(xx-x1)/S1)))+200*(exp(-((xx-x3)/S3)^2)))+(A2*(373*(exp(0.5*(S2/T2)^2-(xx-x2)/T2)*erfc(0.707*(S2/T2)-(xx-x2)/S2))+200*(exp(-((xx-x3)/S3)^2)))
    //CurveFitDialog/ End of Equation
    //CurveFitDialog/ Independent Variables 1
    //CurveFitDialog/ xx
    //CurveFitDialog/ S1 --> Set equal to 0.064
    //CurveFitDialog/ T1 --> Set equal to 0.23
    //CurveFitDialog/ S2 --> Set equal to 0.12
    //CurveFitDialog/ T2 --> Set equal to 0.09
    //CurveFitDialog/ x1 --> Set equal to 1.92
    //CurveFitDialog/ x2 --> Set equal to 2.17
    //CurveFitDialog/ Coefficients 2
    //CurveFitDialog/ ww[0] = A1
    //CurveFitDialog/ ww[1] = A2
   
   
    w1 = exp(0.5*(S1/T1)^2-(xx-x1)/T1)  
    w2 = erfc(0.707*(S1/T1-(xx-x1)/S1))
    w3 = 1110
    w4 = exp(0.5*(S2/T2)^2-(xx-x2)/T2)  
    w5 = erfc(0.707*(S2/T2-(xx-x2)/S2))
    w6 = 373
    w7 = 200
    w8 = exp(-((xx-x3)/S3)^2)
    w9 = 200
    w10 = 560
    w11 = ww[0]
    w12 = ww[1]
    w13 = w3*(w1*w2)+w7*w8
    w14 = w6*(w4*w5)+w10*w8
    qq = w11*w13+w12*w14
    return qq

for(i=0; i<k; i+=1)
        String path = "root:ps_2:input"  //directory of my 2D wave
        Display $path
        DFREF dfr = $path
        k=DimSize ($path,1)
    for(i=0; i<k; i+=1)                        
        Display dfr:f35[][i]   
        FuncFit/NTHR=0/Q/N  FitMySpectrum (f35[][i])    //f35 is the input, 2D wave to which fitting is done

                endfor
Right now you are overwriting your A1, A2 and fitdata waves every time FitMySpectrum is called, and FitMySpectrum is called once for every data point in your wave for every iteration of your fit. You need to create those waves outside the for loop and store the fit results after the FuncFit call, but inside the loop. You also seem to start the for loop twice...

You also need to declare the wave you want to fit using Wave MyWave=dfr:xxxxx

December 3, 2017 at 11:59 pm - Permalink

Plus you call your fitfunction from its inside... (VERY rare)
You are missing and "end" after the return command and you need to start a new function afterwards.
You cannot display an entire data folder: Display $path vs f35 (Maybe you want to display an empty graph: that would be a single dispay command.)

Try to get single tasks working one by one. Start with the fit function, add the displaying of the raw data, do manual fitting, store the coefficients, put this in loops, append the fits, and list the coefficients.
HJ

December 4, 2017 at 01:08 am - Permalink

Just to be clear, the fitFunc function is only to compute a model value. It is called by Igor during iterations during the call to FuncFit.

You are trying to add code to the fitting function that belongs in a separate function. You need a driver function that looks up the wave with the data in it and loops over the columns, calling FuncFit. The loop in your code needs to be put into a separate function, and then you would call that function.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

December 4, 2017 at 09:58 am - Permalink

johnweeks wrote:
Just to be clear, the fitFunc function is only to compute a model value. It is called by Igor during iterations during the call to FuncFit.

You are trying to add code to the fitting function that belongs in a separate function. You need a driver function that looks up the wave with the data in it and loops over the columns, calling FuncFit. The loop in your code needs to be put into a separate function, and then you would call that function.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

Thanks Olelytken, HJ, and John for the suggestions. I slightly modified the code to separate fit function and raw spectrum. I can now display raw spectrums in a separate window and I want to call my fit function to fit each of these spectrums. While I am still working on how to call the fit func and append it to the raw spectrum, I would appreciate any insight/comments/help topic to get started.
#pragma rtGlobals=3     // Use modern global access method and strict wave access.
#include <WMBatchCurveFitIM>

Function FitMySpectrum(ww, xx) : FitFunc

    WAVE ww    
    Variable xx
    Variable w1=0,w2=0,w3=0,w4=0,w5=0,w6=0,w7=0,w8=0,w9=0,w10=0,w11=0,w12=0,w13,w14=0,qq=0
    Variable T1=0.24, S1=0.047, T2=0.24, S2=0.047, x1=1.91, x2=1.918, S3=0.26, x3=2.20
    //Variable k
    //Make/O/N=46 A1   //wave to store coefficients A1
    //Make/O/N=46 A2   //wave to store coefficients A2
    //Make/O/N=(41,46) fitdata    //wavet to store results from fitting
                             
    //CurveFitDialog/ These comments were created by the Curve Fitting dialog. Altering them will
    //CurveFitDialog/ make the function less convenient to work with in the Curve Fitting dialog.
    //CurveFitDialog/ Equation:
    //CurveFitDialog/ f(xx) =  (A1*(1110*(exp(0.5*(S1/T1)^2-(xx-x1)/T1))*erfc(0.707*(S1/T1-(xx-x1)/S1)))+200*(exp(-((xx-x3)/S3)^2)))+(A2*(373*(exp(0.5*(S2/T2)^2-(xx-x2)/T2)*erfc(0.707*(S2/T2)-(xx-x2)/S2))+200*(exp(-((xx-x3)/S3)^2)))
    //CurveFitDialog/ End of Equation
    //CurveFitDialog/ Independent Variables 1
    //CurveFitDialog/ xx
    //CurveFitDialog/ S1 --> Set equal to 0.064
    //CurveFitDialog/ T1 --> Set equal to 0.23
    //CurveFitDialog/ S2 --> Set equal to 0.12
    //CurveFitDialog/ T2 --> Set equal to 0.09
    //CurveFitDialog/ x1 --> Set equal to 1.92
    //CurveFitDialog/ x2 --> Set equal to 2.17
    //CurveFitDialog/ Coefficients 2
    //CurveFitDialog/ ww[0] = A1
    //CurveFitDialog/ ww[1] = A2
   
   
    w1 = exp(0.5*(S1/T1)^2-(xx-x1)/T1)  
    w2 = erfc(0.707*(S1/T1-(xx-x1)/S1))
    w3 = 1110
    w4 = exp(0.5*(S2/T2)^2-(xx-x2)/T2)  
    w5 = erfc(0.707*(S2/T2-(xx-x2)/S2))
    w6 = 373
    w7 = 200
    w8 = exp(-((xx-x3)/S3)^2)
    w9 = 200
    w10 = 560
    w11 = ww[0]
    w12 = ww[1]
    w13 = w3*(w1*w2)+w7*w8
    w14 = w6*(w4*w5)+w10*w8
    qq = w11*w13+w12*w14
    return qq
   
   
End

Function RawSpectrumDisplay()
   
    variable i
    for(i=0;i<46;i+=1)
        String path = "root:ps_2"
        Display $path
        DFREF dfr = $path                          
        Display dfr:f35[][i]       //These are the data I want to fit
    endfor
   
End

December 4, 2017 at 10:29 am - Permalink

johnweeks wrote:
Just to be clear, the fitFunc function is only to compute a model value. It is called by Igor during iterations during the call to FuncFit.

You are trying to add code to the fitting function that belongs in a separate function. You need a driver function that looks up the wave with the data in it and loops over the columns, calling FuncFit. The loop in your code needs to be put into a separate function, and then you would call that function.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

Thanks Olelytken, HJ, and John for the suggestions. I slightly modified the code to separate fit function and raw spectrum. I can now display raw spectrums in a separate window and want to call my fit function to fit each of these spectrums. While I am still working on how to call the fit func and append it to the raw spectrum, I would appreciate any insight/comments/help topic to get started.
#pragma rtGlobals=3     // Use modern global access method and strict wave access.
#include <WMBatchCurveFitIM>

Function FitMySpectrum(ww, xx) : FitFunc

    WAVE ww    
    Variable xx
    Variable w1=0,w2=0,w3=0,w4=0,w5=0,w6=0,w7=0,w8=0,w9=0,w10=0,w11=0,w12=0,w13,w14=0,qq=0
    Variable T1=0.24, S1=0.047, T2=0.24, S2=0.047, x1=1.91, x2=1.918, S3=0.26, x3=2.20
    //Variable k
    //Make/O/N=46 A1   //wave to store coefficients A1
    //Make/O/N=46 A2   //wave to store coefficients A2
    //Make/O/N=(41,46) fitdata    //wavet to store results from fitting
                             
    //CurveFitDialog/ These comments were created by the Curve Fitting dialog. Altering them will
    //CurveFitDialog/ make the function less convenient to work with in the Curve Fitting dialog.
    //CurveFitDialog/ Equation:
    //CurveFitDialog/ f(xx) =  (A1*(1110*(exp(0.5*(S1/T1)^2-(xx-x1)/T1))*erfc(0.707*(S1/T1-(xx-x1)/S1)))+200*(exp(-((xx-x3)/S3)^2)))+(A2*(373*(exp(0.5*(S2/T2)^2-(xx-x2)/T2)*erfc(0.707*(S2/T2)-(xx-x2)/S2))+200*(exp(-((xx-x3)/S3)^2)))
    //CurveFitDialog/ End of Equation
    //CurveFitDialog/ Independent Variables 1
    //CurveFitDialog/ xx
    //CurveFitDialog/ S1 --> Set equal to 0.064
    //CurveFitDialog/ T1 --> Set equal to 0.23
    //CurveFitDialog/ S2 --> Set equal to 0.12
    //CurveFitDialog/ T2 --> Set equal to 0.09
    //CurveFitDialog/ x1 --> Set equal to 1.92
    //CurveFitDialog/ x2 --> Set equal to 2.17
    //CurveFitDialog/ Coefficients 2
    //CurveFitDialog/ ww[0] = A1
    //CurveFitDialog/ ww[1] = A2
   
   
    w1 = exp(0.5*(S1/T1)^2-(xx-x1)/T1)  
    w2 = erfc(0.707*(S1/T1-(xx-x1)/S1))
    w3 = 1110
    w4 = exp(0.5*(S2/T2)^2-(xx-x2)/T2)  
    w5 = erfc(0.707*(S2/T2-(xx-x2)/S2))
    w6 = 373
    w7 = 200
    w8 = exp(-((xx-x3)/S3)^2)
    w9 = 200
    w10 = 560
    w11 = ww[0]
    w12 = ww[1]
    w13 = w3*(w1*w2)+w7*w8
    w14 = w6*(w4*w5)+w10*w8
    qq = w11*w13+w12*w14
    return qq
   
   
End

Function RawSpectrumDisplay()
   
    variable i
    for(i=0;i<46;i+=1)
        String path = "root:ps_2"
        Display $path
        DFREF dfr = $path                          
        Display dfr:f35[][i]       //These are the data I want to fit
    endfor
   
End

December 4, 2017 at 10:26 am - Permalink

 

johnweeks wrote:
Just to be clear, the fitFunc function is only to compute a model value. It is called by Igor during iterations during the call to FuncFit.

You are trying to add code to the fitting function that belongs in a separate function. You need a driver function that looks up the wave with the data in it and loops over the columns, calling FuncFit. The loop in your code needs to be put into a separate function, and then you would call that function.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

I tried to make the use of driver function within which I looped over the columns of my data. Calling the driver function from the command line just displays the raw data however; I can see the fitting coefficients for the last spectrum being displayed on the command panel. Also I copied the second last line of my code form the help menu and do not understand what ''/D /STRC=" is referring to. I would be thankful for any suggestion on how to get the fits appended over the raw spectrum.

 

#pragma rtGlobals=3     // Use modern global access method and strict wave access.
#include <WMBatchCurveFitIM>

Function FitMySpectrum(ww, xx) : FitFunc

    WAVE ww    
    Variable xx
    Variable w1=0,w2=0,w3=0,w4=0,w5=0,w6=0,w7=0,w8=0,w9=0,w10=0,w11=0,w12=0,w13,w14=0,qq=0
    Variable T1=0.24, S1=0.047, T2=0.24, S2=0.047, x1=1.91, x2=1.918, S3=0.26, x3=2.20
    //Variable i, k
    //Make/O/N=46 A1   //wave to store coefficients A1
    //Make/O/N=46 A2   //wave to store coefficients A2
    //Make/O/N=(41,46) fitdata    //wavet to store results from fitting
                             
    //CurveFitDialog/ These comments were created by the Curve Fitting dialog. Altering them will
    //CurveFitDialog/ make the function less convenient to work with in the Curve Fitting dialog.
    //CurveFitDialog/ Equation:
    //CurveFitDialog/ f(xx) =  (A1*(1110*(exp(0.5*(S1/T1)^2-(xx-x1)/T1))*erfc(0.707*(S1/T1-(xx-x1)/S1)))+200*(exp(-((xx-x3)/S3)^2)))+(A2*(373*(exp(0.5*(S2/T2)^2-(xx-x2)/T2)*erfc(0.707*(S2/T2)-(xx-x2)/S2))+200*(exp(-((xx-x3)/S3)^2)))
    //CurveFitDialog/ End of Equation
    //CurveFitDialog/ Independent Variables 1
    //CurveFitDialog/ xx
    //CurveFitDialog/ S1 --> Set equal to 0.064
    //CurveFitDialog/ T1 --> Set equal to 0.23
    //CurveFitDialog/ S2 --> Set equal to 0.12
    //CurveFitDialog/ T2 --> Set equal to 0.09
    //CurveFitDialog/ x1 --> Set equal to 1.92
    //CurveFitDialog/ x2 --> Set equal to 2.17
    //CurveFitDialog/ Coefficients 2
    //CurveFitDialog/ ww[0] = A1
    //CurveFitDialog/ ww[1] = A2
   
   
    w1 = exp(0.5*(S1/T1)^2-(xx-x1)/T1)  
    w2 = erfc(0.707*(S1/T1-(xx-x1)/S1))
    w3 = 1110
    w4 = exp(0.5*(S2/T2)^2-(xx-x2)/T2)  
    w5 = erfc(0.707*(S2/T2-(xx-x2)/S2))
    w6 = 373
    w7 = 200
    w8 = exp(-((xx-x3)/S3)^2)
    w9 = 200
    w10 = 560
    w11 = ww[0]
    w12 = ww[1]
    w13 = w3*(w1*w2)+w7*w8
    w14 = w6*(w4*w5)+w10*w8
    qq = w11*w13+w12*w14
    return qq
   
   
End

Function RawSpectrumDisplay()
   
    variable i
    for(i=0;i<46;i+=1)
        String path = "root:ps_2"
        Display $path
        DFREF dfr = $path                          
        Display dfr:f35[][i]       //These are the data I want to fit
    endfor
   
End

// The structure definition
Structure EMGFitStruct
    Wave cw // required coefficient wave that should be guessed initially
    wave yw  // y value which should come from f35[][i]
    wave xw   // x values already scaled in f35[][i]
    variable j
EndStructure
 
// The driver function that calls Func  Fit:
Function EMGFitDriver(cw,yw,xw)
   
    //FUNCREF RawSpectrumDisplay
    Wave cw // coefficient waves
    Wave yw
    Wave xw
    variable i
    for(i=0;i<46;i+=1)
        String path = "root:ps_2"
        Display $path
        DFREF dfr = $path                          
        Display dfr:f35[][i]       //These are the data I want to fit
    endfor
    STRUCT EMGFitStruct f35
    FuncFit FitMySpectrum, cw, yw /X=xw /D /STRC=f35
END

 

 

 

December 4, 2017 at 12:33 pm - Permalink

So far, I made a little progress but still far from what I need. The fitting appears only for the last raw spectrum displayed. I tried to run for loop for the fitting function and also tried some changes in the last function but no luck. I would appreciate very much for the help on how to include the fit for every displayed raw spectrum. Here is the revised code.
#pragma rtGlobals=3     // Use modern global access method and strict wave access.
#include <WMBatchCurveFitIM>

// The structure definition
Structure EMGFitStruct
    Wave cw // required coefficient wave that should be guessed initially
    wave yw  // y value which should come from f35[][i]
    wave xw   // x values already scaled in f35[][i]
    //variable xx, qq
    //wave ww
    wave inputw
EndStructure

Function FitMySpectrum(ww,xx) : FitFunc
    //STRUCT EMGFitStruct &s
    //variable xx
    //variable qq
   
    WAVE ww
    //wave qq
    Variable xx
    Variable w1=0,w2=0,w3=0,w4=0,w5=0,w6=0,w7=0,w8=0,w9=0,w10=0,w11=0,w12=0,w13,w14=0,qq=0
    Variable T1=0.24, S1=0.047, T2=0.24, S2=0.047, x1=1.91, x2=1.918, S3=0.26, x3=2.20
   
    //Variable i, k
    //Make/O/N=46 A1   //wave to store coefficients A1
    //Make/O/N=46 A2   //wave to store coefficients A2
    //Make/O/N=(41,46) qq    //wavet to store results from fitting
                             
    //CurveFitDialog/ These comments were created by the Curve Fitting dialog. Altering them will
    //CurveFitDialog/ make the function less convenient to work with in the Curve Fitting dialog.
    //CurveFitDialog/ Equation:
    //CurveFitDialog/ f(xx) =  (A1*(1110*(exp(0.5*(S1/T1)^2-(xx-x1)/T1))*erfc(0.707*(S1/T1-(xx-x1)/S1)))+200*(exp(-((xx-x3)/S3)^2)))+(A2*(373*(exp(0.5*(S2/T2)^2-(xx-x2)/T2)*erfc(0.707*(S2/T2)-(xx-x2)/S2))+200*(exp(-((xx-x3)/S3)^2)))
    //CurveFitDialog/ End of Equation
    //CurveFitDialog/ Independent Variables 1
    //CurveFitDialog/ xx
    //CurveFitDialog/ S1 --> Set equal to 0.064
    //CurveFitDialog/ T1 --> Set equal to 0.23
    //CurveFitDialog/ S2 --> Set equal to 0.12
    //CurveFitDialog/ T2 --> Set equal to 0.09
    //CurveFitDialog/ x1 --> Set equal to 1.92
    //CurveFitDialog/ x2 --> Set equal to 2.17
    //CurveFitDialog/ Coefficients 2
    //CurveFitDialog/ ww[0] = A1
    //CurveFitDialog/ ww[1] = A2
   
   
    w1 = exp(0.5*(S1/T1)^2-(xx-x1)/T1)  
    w2 = erfc(0.707*(S1/T1-(xx-x1)/S1))
    w3 = 1110
    w4 = exp(0.5*(S2/T2)^2-(xx-x2)/T2)  
    w5 = erfc(0.707*(S2/T2-(xx-x2)/S2))
    w6 = 373
    w7 = 200
    w8 = exp(-((xx-x3)/S3)^2)
    w9 = 200
    w10 = 560
    w11 = ww[0]
    w12 = ww[1]
    w13 = w3*(w1*w2)+w7*w8
    w14 = w6*(w4*w5)+w10*w8
    qq = w11*w13+w12*w14
    return qq
    //variable i
    //for(i=0;i<2;i+=1)
    //qq=qq[i]
    //endfor
    //return qq

   
End

Function RawSpectrumDisplay (coef, f35)
    wave coef
    wave f35
    variable xx
    STRUCT EMGFitStruct fs
    variable i
    for(i=0;i<2;i+=1)
        String path = "root:ps_2"
        Display $path
        DFREF dfr = $path                          
        Display dfr:f35[][i]       //These are the data I want to fit
   
    wave fs.inputw=dfr:f35
    endfor
   
End
 
// The driver function that calls Func  Fit:
Function EMGFitDriver(W_coef,f35,xx)
   
    //FUNCREF RawSpectrumDisplay
    Wave W_coef // coefficient waves
    Wave f35
    Wave xx
    variable i
    for(i=0;i<3;i+=1)
        String path = "root:ps_2:f35[][i]"
        Display $path
        DFREF dfr = $path                          
        Display dfr:f35[][i]       //These are the data I want to fit
    endfor
    STRUCT EMGFitStruct fs    
    FuncFit FitMySpectrum, W_coef, f35[][i]/D /STRC=fs      //FitMySpectrum is the fitting function
     

December 5, 2017 at 08:08 am - Permalink

Once again, just break it down into easy pieces.

Write a function that only displays the raw data.
Write a fit-function that works with the 'Analysis->Curve Fitting' interface. (I don't think you need the structure feature)

From now on have a close look to the history window:
Fit your data manually.
Manage to store the resulting coefficients manually.
Store the resulting fit wave in a 2d-wave similar to your data.
Add the resulting new fit manually to the display (including coloring etc).

Use the information from the history to modify the displaying function mentioned above. In short words: add these commands within the display loop.

Display the coefficient wave and you should be almost happy.
J

December 5, 2017 at 08:55 am - Permalink

HJDrescher wrote:
Once again, just break it down into easy pieces.

Write a function that only displays the raw data.
Write a fit-function that works with the 'Analysis->Curve Fitting' interface. (I don't think you need the structure feature)

From now on have a close look to the history window:
Fit your data manually.
Manage to store the resulting coefficients manually.
Store the resulting fit wave in a 2d-wave similar to your data.
Add the resulting new fit manually to the display (including coloring etc).

Use the information from the history to modify the displaying function mentioned above. In short words: add these commands within the display loop.

Display the coefficient wave and you should be almost happy.
J

TRK
Thank HJ ! 'Analysis->Curve Fitting' works beautifully and I extract the coefficients from the history/textbox in the graph window. Here I am looking for some automate fitting as I have hundreds of columns in my data (though I am running my loops only for 3 cols in above code for check) and will encounter similar sets of data in future. Initially I tried using simple function but it did not work for me and based on the comments and forum posts, I decided to use structure function. I would be happy and appreciate very much for the help on displaying the fit to every raw spectrums than just the last raw spectum in the loop.
Thanks,
Tika

December 5, 2017 at 09:26 am - Permalink

Well then the next step would be to store your fit in a 2d wave. Make sure that you use the right column.
Once this works, add that command right after the funcfit command in your procedure.

December 5, 2017 at 09:49 am - Permalink

HJDrescher wrote:
Well then the next step would be to store your fit in a 2d wave. Make sure that you use the right column.
Once this works, add that command right after the funcfit command in your procedure.


Once again I need help to extract data for each iteration in FuncFit. I checked FuncFit section in help file and it does not explain how to extract results from FuncFit. I just keep on getting the fit only for the last iteration of my loop; it looks like overwriting the fit. Here is what I did; After FuncFit command line I duplicated the wave to get fit data for each iteration (but this does not seem to duplicate the fit data). Then I created a 2D wave to store the fit data. Here is the segment of the code. Thanks
Function EMGFitDriver(W_coef,f35,xx)
     
    Wave W_coef // coefficient waves in my folder
    Wave f35
    Wave xx
    Wave fit_f35
    //Wave outputw
    variable i,n
    for(i=0;i<10;i+=1)
        String path = "root:ps_2:f35[][i]"
        Display $path
        DFREF dfr = $path                          
        Display dfr:f35[][i]       //These are the data I want to fit
    endfor
    STRUCT EMGFitStruct fs      
    FuncFit FitMySpectrum, W_coef, f35[][i]/D/STRC=fs     //FitMySpectrum is the fitting function
    duplicate/o fit_f35, $("fit"+num2istr(n)+"_"+"f35")    //I am assuming FucnFit returns the fit data as fit_f35 wave by default because this is what I see in my folder
    Make/O/N=(41,46) fitdata
    fitdata = fit_f35
    SetAxis bottom 1.64, 3.2
    //Display outputw[][i]
    AppendToGraph  f35[][i],fit_f35[][i]

December 5, 2017 at 04:05 pm - Permalink

I understand that I am making too many posts for a simple problem which might be annoying to other igor programmers/forum readers, I apologize. As a newbie in programming skills, I just wanted to make sure that I am following the right direction and get some insight on the coding from the experts/experienced igor users. In the previous post, I realized my silly mistake of ending the for loop before its execution and now I fixed that, and am able to append the raw spectrum to fit data and display graph for each iteration. In the next step, I am trying to save fit coefficients and fit data of each iteration in a 2D wave form. I would appreciate any help on this part. Here is the segment of the code.
Function EMGFitDriver(W_coef,f35,xx)
     
    Wave W_coef // coefficient waves
    Wave f35
    Wave xx
    Make/O/N = (46)fit_f35
    //Wave outputw
    variable i,n
    for(i=0;i<2;i+=1)
        String path = "root:ps_2:f35[][i]"
        Display $path
        DFREF dfr = $path                          
        Display dfr:f35[][i]       //These are the data I want to fit
   
    STRUCT EMGFitStruct fs      
    FuncFit FitMySpectrum, W_coef, f35[][i]/D/STRC=fs     //FitMySpectrum is the fitting function
    duplicate/o fit_f35, $("fit"+num2istr(n)+"_"+"f35[][i]")
    display fit_f35[][i], $("fit"+num2istr(i)+"_"+"f35[][i]")
    //wave fitdata=$("fit"+num2istr(n)+"_"+"f35")
    Make/O/N=(41,46) fitdata
    wave fitdata=$("fit"+num2istr(0)+"_"+"f35[][i]")
    SetAxis bottom 1.64, 3.2
    endfor
End

December 5, 2017 at 08:41 pm - Permalink

Once again: do it manually first. For three data sets is enough.

Run your code in the command window and see what happens.
I guess you will have more windows created than you actually need.
I guess your storage wave will be overwritten.

Ill guide you though the development. I assume there are things in the code (e.g., "/TN=") that you don't understand. This is absolutely fine. Consult the manual and if you have additional questions, please feel free to ask.

Let's start.

We do not want to mess with the current data folder (DF): Hence, we save the current one, switch to our data, and at the end of the function we switch back.
I assume you want one window per data set (column). Your initial displaying function should actually look like this:
Function ProcessData(DataFolderName, DataName)
    String DataFolderName, DataName    
    // This is the full path to your data in two pieces as text. It could also be passed as a wave reference, but this makes the code a little more complicated
    Variable i
    DFRef OldDF=GetDataFolderDFR()  //Store the current DF
    SetDataFolder $DataFolderName   // Set it to our data
    Wave Data=$DataName
    For (i=0;i<Dimsize(Data,1);i+=1)
        Display /K=1 /W=(10*i,10*i+40, 10*i+440, 10*i+40+210) Data[][i]/TN=$(DataName+"_"+Num2str(i))
        ModifyGraph mode(test_2)=2, lsize(test_2)=2
    EndFor
    SetDataFolder OldDF  // Switch DF back
End

Have a look at the manual what the K and W flags are doing.
you should be able to call it like this:
ProcessData("root:ps_2:input:", "f35")

The next step would be to include the fitting function. There is no need for storing in the fit function. All it does is getting one x value and calculating one y value. For each point in your data. In each iteration. You see, it gets called many times. You want the data after all of this is finished. In the meantime, avoid any overhead: Remove all 'Make' command from the fit-function. Variables i & k are also not used - remove this line as well.

Now you want to prepare for storing something. Your fits will share the same data structure as your raw data. Why not use this property?
Duplicate /O Data $("fit_"+DataName")
and show this new wave to the compiler
Wave Fits=$("fit_"+DataName)
Also we need to create the waves for the coefficients; their size is the same as numbers of columns in your data:
Make /O /N=(DimSize(Data,1)) A1, A2
Since this is related to 2d-Data, these commands must be outside the loop:
Function ProcessData(DataFolderName, DataName)
String DataFolderName, DataName    
Variable i
    DFRef OldDF=GetDataFolderDFR()
    SetDataFolder $DataFolderName
    Wave Data=$DataName
    Duplicate /O Data $("fit_"+DataName)  // Replicate the data structure incl. wavc scaling units etc.
    Wave Fits=$("fit_"+DataName)
    Make /O /N=(DimSize(Data,1)) A1, A2  // Prepare the coefficient storage
    For (i=0;i<Dimsize(Data,1);i+=1)
        Display /K=1 /W=(10*i,10*i+40, 10*i+440, 10*i+40+210) Data[][i]/TN=$(DataName+"_"+Num2str(i))
        ModifyGraph mode(test_2)=2, lsize(test_2)=2
    Endfor
    SetDataFolder OldDF
End

Now we can add the fitting. You don't need the data structure. A temporary (/FREE) coefficient wave with initial guesses is sufficient. Also prepare a temporary wave for the fit outside the loop.
    Make /FREE /D /N=2 w_coef={1,2}
    Make /FREE /N=(Dimsize(Data,0)) Fit

And add the fitting in the loop:
    FuncFit /Q FitMySpectrum W_coef Data[][i] /D=Fit


Function ProcessData(DataFolderName, DataName)
String DataFolderName, DataName    
Variable i
    DFRef OldDF=GetDataFolderDFR()
    SetDataFolder $DataFolderName
    Wave Data=$DataName
    Duplicate /O Data $("fit_"+DataName)
    Wave Fits=$("fit_"+DataName)
    Make /O /N=(DimSize(Data,1)) A1, A2
    Make /FREE /D /N=2 w_coef={1,2}
    Make /FREE /N=(Dimsize(Data,0)) Fit
    For (i=0;i<Dimsize(Data,1);i+=1)
        Display /K=1 /W=(10*i,10*i+40, 10*i+440, 10*i+40+210) Data[][i]/TN=$(DataName+"_"+Num2str(i))
        ModifyGraph mode($(DataName+"_"+Num2str(i)))=2, lsize($(DataName+"_"+Num2str(i)))=2
        FuncFit /Q FitMySpectrum W_coef Data[][i] /D=Fit
    Endfor
    SetDataFolder OldDF
End

If you test this function, you will realize that there is no fit displayed in the windows. This is ok for now. Let's store our results first.
        Fits[][i]=Fit[p]
        A1[i]=w_coef[0]
        A2[i]=w_coef[1]

Certainly this needs to be done right after the fitting.
Next, we nicely display the resulting fit within the loop
        AppendToGraph Fits[][i]/TN=$(DataName+"_Fit_"+Num2str(i))
        ModifyGraph rgb($(DataName+"_Fit_"+Num2str(i)))=(0,0,0)

Finally we add some output of the coefficeints
    Edit /K=1 /W=(400,40,800,240) A1 A2
    Display /K=1 /W=(400,310,800,510)  A1
    Display /K=1 /W=(400,550,800,750) A2


Altogether it should look like this:
Function ProcessData(DataFolderName, DataName)
String DataFolderName, DataName    
Variable i
    DFRef OldDF=GetDataFolderDFR()
    SetDataFolder $DataFolderName
    Wave Data=$DataName
    Duplicate /O Data $("fit_"+DataName)
    Wave Fits=$("fit_"+DataName)
    Make /O /N=(DimSize(Data,1)) A1, A2
    Make /FREE /D /N=2 w_coef={1,2}
    Make /FREE /N=(Dimsize(Data,0)) Fit
    For (i=0;i<Dimsize(Data,1);i+=1)
        Display /K=1 /W=(10*i,10*i+40, 10*i+440, 10*i+40+210) Data[][i]/TN=$(DataName+"_"+Num2str(i))
        ModifyGraph mode($(DataName+"_"+Num2str(i)))=2, lsize($(DataName+"_"+Num2str(i)))=2
        FuncFit /Q FitMySpectrum W_coef Data[][i] /D=Fit
        Fits[][i]=Fit[p]
        A1[i]=w_coef[0]
        A2[i]=w_coef[1]
        AppendToGraph Fits[][i]/TN=$(DataName+"_Fit_"+Num2str(i))
        ModifyGraph rgb($(DataName+"_Fit_"+Num2str(i)))=(0,0,0)
    Endfor
    Edit /K=1 /W=(400,40,800,240) A1 A2
    Display /K=1 /W=(400,310,800,510)  A1
    Display /K=1 /W=(400,550,800,750) A2
    SetDataFolder OldDF
End


A final remark: FitMySpectrum is a name that suits for this forum. In Igor, it should get a more meaningful name, e.g. NeuronResponse or DoniachSunjic.

Hope this helps,
HJ

December 6, 2017 at 01:16 am - Permalink

HJDrescher wrote:
Once again: do it manually first. For three data sets is enough.

Run your code in the command window and see what happens.
I guess you will have more windows created than you actually need.
I guess your storage wave will be overwritten.

Ill guide you though the development. I assume there are things in the code (e.g., "/TN=") that you don't understand. This is absolutely fine. Consult the manual and if you have additional questions, please feel free to ask.

Let's start.

We do not want to mess with the current data folder (DF): Hence, we save the current one, switch to our data, and at the end of the function we switch back.
I assume you want one window per data set (column). Your initial displaying function should actually look like this:
Function ProcessData(DataFolderName, DataName)
    String DataFolderName, DataName    
    // This is the full path to your data in two pieces as text. It could also be passed as a wave reference, but this makes the code a little more complicated
    Variable i
    DFRef OldDF=GetDataFolderDFR()  //Store the current DF
    SetDataFolder $DataFolderName   // Set it to our data
    Wave Data=$DataName
    For (i=0;i<Dimsize(Data,1);i+=1)
        Display /K=1 /W=(10*i,10*i+40, 10*i+440, 10*i+40+210) Data[][i]/TN=$(DataName+"_"+Num2str(i))
        ModifyGraph mode(test_2)=2, lsize(test_2)=2
    EndFor
    SetDataFolder OldDF  // Switch DF back
End

Have a look at the manual what the K and W flags are doing.
you should be able to call it like this:
ProcessData("root:ps_2:input:", "f35")

The next step would be to include the fitting function. There is no need for storing in the fit function. All it does is getting one x value and calculating one y value. For each point in your data. In each iteration. You see, it gets called many times. You want the data after all of this is finished. In the meantime, avoid any overhead: Remove all 'Make' command from the fit-function. Variables i & k are also not used - remove this line as well.

Now you want to prepare for storing something. Your fits will share the same data structure as your raw data. Why not use this property?
Duplicate /O Data $("fit_"+DataName")
and show this new wave to the compiler
Wave Fits=$("fit_"+DataName)
Also we need to create the waves for the coefficients; their size is the same as numbers of columns in your data:
Make /O /N=(DimSize(Data,1)) A1, A2
Since this is related to 2d-Data, these commands must be outside the loop:
Function ProcessData(DataFolderName, DataName)
String DataFolderName, DataName    
Variable i
    DFRef OldDF=GetDataFolderDFR()
    SetDataFolder $DataFolderName
    Wave Data=$DataName
    Duplicate /O Data $("fit_"+DataName)  // Replicate the data structure incl. wavc scaling units etc.
    Wave Fits=$("fit_"+DataName)
    Make /O /N=(DimSize(Data,1)) A1, A2  // Prepare the coefficient storage
    For (i=0;i<Dimsize(Data,1);i+=1)
        Display /K=1 /W=(10*i,10*i+40, 10*i+440, 10*i+40+210) Data[][i]/TN=$(DataName+"_"+Num2str(i))
        ModifyGraph mode(test_2)=2, lsize(test_2)=2
    Endfor
    SetDataFolder OldDF
End

Now we can add the fitting. You don't need the data structure. A temporary (/FREE) coefficient wave with initial guesses is sufficient. Also prepare a temporary wave for the fit outside the loop.
    Make /FREE /D /N=2 w_coef={1,2}
    Make /FREE /N=(Dimsize(Data,0)) Fit

And add the fitting in the loop:
    FuncFit /Q FitMySpectrum W_coef Data[][i] /D=Fit


Function ProcessData(DataFolderName, DataName)
String DataFolderName, DataName    
Variable i
    DFRef OldDF=GetDataFolderDFR()
    SetDataFolder $DataFolderName
    Wave Data=$DataName
    Duplicate /O Data $("fit_"+DataName)
    Wave Fits=$("fit_"+DataName)
    Make /O /N=(DimSize(Data,1)) A1, A2
    Make /FREE /D /N=2 w_coef={1,2}
    Make /FREE /N=(Dimsize(Data,0)) Fit
    For (i=0;i<Dimsize(Data,1);i+=1)
        Display /K=1 /W=(10*i,10*i+40, 10*i+440, 10*i+40+210) Data[][i]/TN=$(DataName+"_"+Num2str(i))
        ModifyGraph mode($(DataName+"_"+Num2str(i)))=2, lsize($(DataName+"_"+Num2str(i)))=2
        FuncFit /Q FitMySpectrum W_coef Data[][i] /D=Fit
    Endfor
    SetDataFolder OldDF
End

If you test this function, you will realize that there is no fit displayed in the windows. This is ok for now. Let's store our results first.
        Fits[][i]=Fit[p]
        A1[i]=w_coef[0]
        A2[i]=w_coef[1]

Certainly this needs to be done right after the fitting.
Next, we nicely display the resulting fit within the loop
        AppendToGraph Fits[][i]/TN=$(DataName+"_Fit_"+Num2str(i))
        ModifyGraph rgb($(DataName+"_Fit_"+Num2str(i)))=(0,0,0)

Finally we add some output of the coefficeints
    Edit /K=1 /W=(400,40,800,240) A1 A2
    Display /K=1 /W=(400,310,800,510)  A1
    Display /K=1 /W=(400,550,800,750) A2


Altogether it should look like this:
Function ProcessData(DataFolderName, DataName)
String DataFolderName, DataName    
Variable i
    DFRef OldDF=GetDataFolderDFR()
    SetDataFolder $DataFolderName
    Wave Data=$DataName
    Duplicate /O Data $("fit_"+DataName)
    Wave Fits=$("fit_"+DataName)
    Make /O /N=(DimSize(Data,1)) A1, A2
    Make /FREE /D /N=2 w_coef={1,2}
    Make /FREE /N=(Dimsize(Data,0)) Fit
    For (i=0;i<Dimsize(Data,1);i+=1)
        Display /K=1 /W=(10*i,10*i+40, 10*i+440, 10*i+40+210) Data[][i]/TN=$(DataName+"_"+Num2str(i))
        ModifyGraph mode($(DataName+"_"+Num2str(i)))=2, lsize($(DataName+"_"+Num2str(i)))=2
        FuncFit /Q FitMySpectrum W_coef Data[][i] /D=Fit
        Fits[][i]=Fit[p]
        A1[i]=w_coef[0]
        A2[i]=w_coef[1]
        AppendToGraph Fits[][i]/TN=$(DataName+"_Fit_"+Num2str(i))
        ModifyGraph rgb($(DataName+"_Fit_"+Num2str(i)))=(0,0,0)
    Endfor
    Edit /K=1 /W=(400,40,800,240) A1 A2
    Display /K=1 /W=(400,310,800,510)  A1
    Display /K=1 /W=(400,550,800,750) A2
    SetDataFolder OldDF
End


A final remark: FitMySpectrum is a name that suits for this forum. In Igor, it should get a more meaningful name, e.g. NeuronResponse or DoniachSunjic.

Hope this helps,
HJ

Thank you HJ for step by step development towards the goal. This is what I was looking for exactly. I appreciate your time and effort. It is so good to see how flags after the display command can kill the window (/K=1) with no dialog, determine the location and size of the graph (/w) and provide tracename based on the parent wave(/TN).
At the end, I included a residual wave to visualize how good the fit is. Finally, have a couple of quick questions. I checked that the script "Residuals=Data-Fits" works as it calculates after all the waves are collected. At the same time why "Residuals[][i]=Data[i]-Fits[i] or Residuals[][p]=Data[p]-Fits[p]" does not work inside the loop because I was assuming that the subtraction from the raw data takes place for each iteration and is stored in Residuals.
Next, I know that I can chose not to display the graph window by disabling the display command. I noticed there is /k=3 to hide the window. Other than these is there any quick way to kill the graph window rather than doing it manually?
Once again, thank you very much !
Here is how I tried to calculate Residuals,
Function ProcessData(DataFolderName, DataName)
String DataFolderName, DataName    
Variable i
    DFRef OldDF=GetDataFolderDFR()
    SetDataFolder $DataFolderName     //identifies the path for the current data folder
    Wave Data=$DataName                  // wavename to be executed
    Duplicate /O Data $("fit_"+DataName)
    Wave Fits=$("fit_"+DataName)
    Make /O /N=(DimSize(Data,1)) A1, A2
    Make /FREE/D /N=2 w_coef={1,2}
    Make /FREE/N=(Dimsize(Data,0)) Fit
    Duplicate /O Data $("Residual_"+DataName)  //////////////
    Wave Residuals=$("Residual_"+DataName)   //////////////
    variable k
    k=Dimsize(Data,1)
    For (i=0;i<Dimsize(Data,1);i+=1)
        Display /K=1 /W=(10*i,10*i+40, 10*i+440, 10*i+40+210) Data[][i]/TN=$(DataName+"_"+Num2str(i))
        ModifyGraph mode($(DataName+"_"+Num2str(i)))=2, lsize($(DataName+"_"+Num2str(i)))=2
        FuncFit /Q FitMySpectrum W_coef Data[][i] /D=Fit
        Fits[][i]=Fit[p]
        Residuals=Data-Fits               ////////////////residuals are calculated
        //Residuals[][i]=Data[i]-Fits[i]  
        //Residuals[p][q]=Data[p][q]-Fits[p][q]     //this does not work
        A1[i]=w_coef[0]
        A2[i]=w_coef[1]
        AppendToGraph Fits[][i]/TN=$(DataName+"_Fit_"+Num2str(i))
        ModifyGraph rgb($(DataName+"_Fit_"+Num2str(i)))=(0,0,0)
    Endfor
    //Residuals=Data-Fits                    ////////////////////////
    Edit /K=1 /W=(400,100,800,240) A1 A2
    Display /K=1 /W=(400,310,800,510)  A1
    Display /K=1 /W=(400,550,800,750) A2
    SetDataFolder OldDF
End

<pre><code class="language-igor">

December 6, 2017 at 05:51 pm - Permalink

Small remark: Please don't quote every last response, the thread gets longish. "Post reply" is fine.

Calculating the residuals at the end of the procedure should work that way you wrote it.
Inside the loop you have to pay attention to the dimensions of the waves ( number of [...] pairs ), indices (i), automatic loop indices (p, q, x, y, z etc) and "do-it-for-all-points-in-that-dimension" ( [] )

The right syntax would be
Residuals[][i]=Data[p][i]-Fit[p]

You want to calculate for all rows (which correspond to index p) in column i in Residuals (this is the [][i] part) the difference between row p in column p of data and the p-th row in Fit.
Please read the manual about wave assignment:
displayhelptopic "Waveform Arithmetic and Assignment"
displayhelptopic "Multidimensional Wave Assignment"

(Copy these commands to the command window and execute them)

Essentially, KillWindow does the job of closing a window. Sadly not a list of windows. So you'd need to write a procedure for this.
function Killwins(WindowList)
string WindowList
variable i
    for (i=0;i<itemsinlist(WindowList);i+=1)
        Killwindow $stringfromlist(i,WindowList)
    endfor
end

To kill all graphs run
Killwins(winlist("*",";","WIN:1"))
and to kill all tables
Killwins(winlist("*",";","WIN:2"))

However, it might not be necessary to kill the windows, e.g, if you only want to rerun the fitting. In this case, it might be advisable to separate the processing function into smaller fragments that share the same task. Examples would be creation of data structure, displaying, fitting, summarizing, and cleanup. The same holds, if you want to recreate the graphs later without new fitting. I can give you a step by step example later.

I added a little extra to the processing function:
Function ProcessData(DataFolderName, DataName)
String DataFolderName, DataName    
Variable i
    DFRef OldDF=GetDataFolderDFR()
    SetDataFolder $DataFolderName
    Wave Data=$DataName
    Duplicate /O Data $("fit_"+DataName), $("res_"+DataName) // You can duplicate to several destination waves; Keep the names short for automatic extension
    Wave Residuals=$("res_"+DataName)
    Wave Fits=$("fit_"+DataName)
    Make /O /N=(DimSize(Data,1)) A1, A2
    Make /FREE /D /N=2 w_coef={1,2}
    Make /FREE /N=(Dimsize(Data,0)) Fit
    For (i=0;i<Dimsize(Data,1);i+=1)
        Display /K=1 /W=(10*i,10*i+40, 10*i+440, 10*i+40+210) Data[][i]/TN=$(DataName+"_"+Num2str(i)) as DataName+"_"+num2str(i)
        FuncFit /Q FitMySpectrum W_coef Data[][i] /D=Fit
        Fits[][i]=Fit[p]
        Residuals[][i]=Data[p][i]-Fit[p]
        A1[i]=w_coef[0]
        A2[i]=w_coef[1]
        AppendToGraph Fits[][i]/TN=$(DataName+"_Fit_"+Num2str(i))
        AppendToGraph /L=laxRes Residuals[][i]/TN=$(DataName+"_Res_"+Num2str(i))
        ModifyGraph mode($(DataName+"_"+Num2str(i)))=2, lsize($(DataName+"_"+Num2str(i)))=2
        ModifyGraph mode($(DataName+"_Res_"+Num2str(i)))=2, lsize($(DataName+"_Res_"+Num2str(i)))=2
        ModifyGraph freePos(laxRes)={0,bottom}
        ModifyGraph axisEnab(left)={0,0.75}
        ModifyGraph axisEnab(laxRes)={0.8,1}
        ModifyGraph rgb($(DataName+"_Fit_"+Num2str(i)))=(0,0,0)
    Endfor
//  Residuals=Data-Fits
    Edit /K=1 /W=(400,40,800,240) A1 A2
    Display /K=1 /W=(400,310,800,510)  A1
    Display /K=1 /W=(400,550,800,750) A2
    SetDataFolder OldDF
End


HJ

December 7, 2017 at 05:05 am - Permalink

Thank you HJ for guiding step by step through example. Now I can see how indexing in wave assignment works and the technique how multiple tracenames can be replaced by a single name. I noticed that /TN flags for the residuals and fits could be disabled as it has no affect unless we want to change color of the trace. Also thanks for introducing free axes and how traces can be beautifully positioned for the better display. In fact I started enjoying programming in igor much than before and I understand there is a lot more to know. I am keen to know how above function can be processed into fragments if your time permits.
Thank you.

December 7, 2017 at 11:48 pm - Permalink

I think you may be ready to read the Programming help file: DisplayHelpTopic "Programming Overview", in Programming.ihf. There is another help file: Programming Techniques.ihf that has a lot of good advice.

It is hard going. The first time through it won't make a lot of sense. But some will stick and the next time through more of it will make sense. And it helps to have a project like yours to apply your learning to as you read.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

December 8, 2017 at 09:17 am - Permalink

Thank you Johnweeks for the suggestion. I will go through the programming techniques while it may take time to be familiar with. For now, I have a quick question related to scaling of waves.
I have two different 2D waves with same number of rows (energy) but different columns (time). These waves overlap in columns, that is first wave is data from 40ps to -5ps with deltay= -1 ps and second wave is data from 300 ps to -50 ps with deltay = -10 ps. I concatenated the two waves to get a new wave destWave( 46*72) which now has data from 300 to 40 ps (col0 to col26) with deltay = -10 and 39 ps to -5 ps (col27 to col72). Since the concatenated wave takes scaling of first wave in the list, ywave scaling mismatches. I tried to run a do while loop but it returns 'nan' in col scaling and the graph is not displayed. Can we scale a 2D wave with two different deltay values?
Here is the script I tried with.
Function ConcatenateWaves(data2, data1)
    Wave data2, data1
    Wave destWave
    variable k, dy
    //dy =destWave[0] - destWave[1]
    Duplicate /O/R=[][1,*] data1, sub1
    Duplicate /O/R=[][0,26] data2, sub2
    k=Dimsize(destWave,1)
    Concatenate/O/NP=1 {sub2,sub1}, destWave
    //for (k=0;k<26;k+=1)
    do (k<27)            //col 0 to col 26; starty 300, deltay= -10, endy 40;
     dy = 300-10*k     //col 27 to col72, starty39, deltay= -1, endy= -5
    while (k>26)
    dy = 39-1*k
    SetScale/P y, destWave[0], dy, destWave
    //SetAxis/A/R left

    //endfor
End
<pre><code class="language-igor">

December 9, 2017 at 09:43 am - Permalink

No, a wave can have only one scaling for a given dimension. You need to use a 1D auxiliary wave giving the time values for each column. How you make that wave will depend on how you want to use the data. Most likely you just need to use a wave with one time value for each column of the wave. That would appropriate for curve fits, for instance.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

December 11, 2017 at 09:33 am - Permalink

Just to become clear; do you mean that 2D waves need to be changed to 1D wave for the desired scaling? Actually I wanted to see how 2D image changes with time?
Thanks.

December 11, 2017 at 09:59 pm - Permalink

No, I mean that you need a second wave containing the values of time that go with each column in your 2D wave. Once you do that, then you need to keep track of the 1D wave whenever you need those time values.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

December 12, 2017 at 09:16 am - Permalink

If you x-values are not following the equation x=delta*p+offset you need x/y pairs (the second 1D wave John mentioned).
Please read
displayhelptopic "The XY Model of Data"

December 13, 2017 at 12:52 am - Permalink

Back to our step-by-step example (sorry for replying late, I had some quite busy time):

First, we need to identify each command with a logical set of actions.
Function ProcessData(DataFolderName, DataName)
String DataFolderName, DataName    
Variable i
    DFRef OldDF=GetDataFolderDFR()      // Preparation
    SetDataFolder $DataFolderName       // Preparation
    Wave Data=$DataName     // Preparation
    Duplicate /O Data $("fit_"+DataName), $("res_"+DataName)        // Data management
    Wave Residuals=$("res_"+DataName)       // Preparation
    Wave Fits=$("fit_"+DataName)    // Preparation
    Make /O /N=(DimSize(Data,1)) A1, A2 // Data management
    Make /FREE /D /N=2 w_coef={1,2}     // Preparation
    Make /FREE /N=(Dimsize(Data,0)) Fit     // Preparation
    For (i=0;i<Dimsize(Data,1);i+=1)        // Loop
        Display /K=1 /W=(10*i,10*i+40, 10*i+440, 10*i+40+210) Data[][i]/TN=$(DataName+"_"+Num2str(i)) as DataName+"_"+num2str(i)        // Display
        FuncFit /Q FitMySpectrum W_coef Data[][i] /D=Fit        // Fitting
        Fits[][i]=Fit[p]        // Fitting
        Residuals[][i]=Data[p][i]-Fit[p]        // Fitting
        A1[i]=w_coef[0]     // Fitting
        A2[i]=w_coef[1]     // Fitting
        AppendToGraph Fits[][i]/TN=$(DataName+"_Fit_"+Num2str(i))       // Display
        AppendToGraph /L=laxRes Residuals[][i]/TN=$(DataName+"_Res_"+Num2str(i))        // Display
        ModifyGraph mode($(DataName+"_"+Num2str(i)))=2, lsize($(DataName+"_"+Num2str(i)))=2     // Display
        ModifyGraph mode($(DataName+"_Res_"+Num2str(i)))=2, lsize($(DataName+"_Res_"+Num2str(i)))=2     // Display
        ModifyGraph freePos(laxRes)={0,bottom}      // Display
        ModifyGraph axisEnab(left)={0,0.75}     // Display
        ModifyGraph axisEnab(laxRes)={0.8,1}        // Display
        ModifyGraph rgb($(DataName+"_Fit_"+Num2str(i)))=(0,0,0)     // Display
    Endfor      // Loop
    Edit /K=1 /W=(400,40,800,240) A1 A2     // Summary
    Display /K=1 /W=(400,310,800,510)  A1       // Summary
    Display /K=1 /W=(400,550,800,750) A2        // Summary
    SetDataFolder OldDF     // Clean up
End


The next steps are to create a copy of the function for each set, rename them appropriately, and delete parts that don't match. Be careful with loops and preparations! They might appear in several places.
(This example is a little overkill, I know. However, it's good for illustration.)
Function PrepareData(DataFolderName, DataName)
String DataFolderName, DataName    
Variable i
    DFRef OldDF=GetDataFolderDFR()      // Preparation
    SetDataFolder $DataFolderName       // Preparation
    Wave Data=$DataName     // Preparation
    Duplicate /O Data $("fit_"+DataName), $("res_"+DataName)        // Data management
    Make /O /N=(DimSize(Data,1)) $(DataName+"_A1"),  $(DataName+"_A2")  // Data management
    SetDataFolder OldDF     // Clean up
End

Function FitData(DataFolderName, DataName)
String DataFolderName, DataName    
Variable i
    DFRef OldDF=GetDataFolderDFR()      // Preparation
    SetDataFolder $DataFolderName       // Preparation
    Wave Data=$DataName     // Preparation
    Wave Residuals=$("res_"+DataName)       // Preparation
    Wave Fits=$("fit_"+DataName)    // Preparation
    Wave A1= $(DataName+"_A1"), A2= $(DataName+"_A2")   // Preparation  <--- This has changed !
    Make /FREE /D /N=2 w_coef={1,2}     // Preparation
    Make /FREE /N=(Dimsize(Data,0)) Fit     // Preparation
    For (i=0;i<Dimsize(Data,1);i+=1)        // Loop
        FuncFit /Q FitMySpectrum W_coef Data[][i] /D=Fit        // Fitting
        Fits[][i]=Fit[p]        // Fitting
        Residuals[][i]=Data[p][i]-Fit[p]        // Fitting
        A1[i]=w_coef[0]     // Fitting
        A2[i]=w_coef[1]     // Fitting
    Endfor      // Loop
    SetDataFolder OldDF     // Clean up
End

Function DisplayData(DataFolderName, DataName)
String DataFolderName, DataName    
Variable i
    DFRef OldDF=GetDataFolderDFR()      // Preparation
    SetDataFolder $DataFolderName       // Preparation
    Wave Data=$DataName     // Preparation
    Wave Residuals=$("res_"+DataName)       // Preparation
    Wave Fits=$("fit_"+DataName)    // Preparation
    For (i=0;i<Dimsize(Data,1);i+=1)        // Loop
        Display /K=1 /W=(10*i,10*i+40, 10*i+440, 10*i+40+210) Data[][i]/TN=$(DataName+"_"+Num2str(i)) as DataName+"_"+num2str(i)        // Display
        AppendToGraph Fits[][i]/TN=$(DataName+"_Fit_"+Num2str(i))       // Display
        AppendToGraph /L=laxRes Residuals[][i]/TN=$(DataName+"_Res_"+Num2str(i))        // Display
        ModifyGraph mode($(DataName+"_"+Num2str(i)))=2, lsize($(DataName+"_"+Num2str(i)))=2     // Display
        ModifyGraph mode($(DataName+"_Res_"+Num2str(i)))=2, lsize($(DataName+"_Res_"+Num2str(i)))=2     // Display
        ModifyGraph freePos(laxRes)={0,bottom}      // Display
        ModifyGraph axisEnab(left)={0,0.75}     // Display
        ModifyGraph axisEnab(laxRes)={0.8,1}        // Display
        ModifyGraph rgb($(DataName+"_Fit_"+Num2str(i)))=(0,0,0)     // Display
    Endfor      // Loop
    SetDataFolder OldDF     // Clean up
End

Function SummaryData(DataFolderName, DataName)
String DataFolderName, DataName
Variable i
    DFRef OldDF=GetDataFolderDFR()      // Preparation
    SetDataFolder $DataFolderName       // Preparation
    Wave Data=$DataName     // Preparation
    Duplicate /O Data $("fit_"+DataName), $("res_"+DataName)        // Data management
    Wave Residuals=$("res_"+DataName)       // Preparation
    Wave Fits=$("fit_"+DataName)    // Preparation
    Wave A1= $(DataName+"_A1"), A2= $(DataName+"_A2")   // Preparation  <--- This has changed from creation to reference
    Edit /K=1 /W=(400,40,800,240) A1 A2     // Summary
    Display /K=1 /W=(400,310,800,510)  A1       // Summary
    Display /K=1 /W=(400,550,800,750) A2        // Summary
    SetDataFolder OldDF     // Clean up
End


Now you have to run the functions one by one
PrepareData("root:sdf:", "test")
FitData("root:sdf:", "test")
DisplayData("root:sdf:", "test")
SummaryData("root:sdf:", "test")


Note: If you run one of the latter functions on a new set of data, it will fail. It expects the prepared waves to exist, which is not the case for a new data set. Here, some data structure checking would be required to protect the data from the user. Have a look in the manual and read about if ... else ... endif-structures and the exists / waveexists functions.

You might have realized that the original function is gone at this point. It was so nice, that it did all of the work with one command. Let's bring it back:
Function ProcessData(DataFolderName, DataName)
String DataFolderName, DataName
    PrepareData(DataFolderName, DataName)
    FitData(DataFolderName, DataName)
    DisplayData(DataFolderName, DataName)
    SummaryData(DataFolderName, DataName)
End

It simply takes your input and calls the other functions. What is it good for? We had the same result using one more complex function.
On the one hand, it gives the programmer a better view on his code. Once the simple functions reach a couple of hundred lines you are happy that it's not in the thousands.
On the other hand, we can re-run single steps!
Assume you did a full run on the data, modified the graphs, added some manual annotation, ..... and you figured out that there was a small typo in the fit function. You don't want to kill all the nice figures. Actually, all you need is to re-run the fit and check your remarks in the graph. This is possible now. The data structure is prepared from the last run. You can call the fitting by itself. Igor updates the graphs as the corresponding waves are altered.

As a next logical step, we want to kill (or hide but this takes some more checking) the fit result windows. In order to do this, we should give the windows names:
Display /N=$(DataName+"_"+Num2str(i))) /K=1 /W=(10*i,10*i+40, 10*i+440, 10*i+40+210) Data[][i]/TN=$(DataName+"_"+Num2str(i)) as DataName+"_"+num2str(i)        // in DisplayData

and prepare a function that kills the windows related to a data set (based on that name):
Function KillFitWindows(DataFolderName, DataName)
String DataFolderName, DataName
Variable i
    Wave Data=$(DataFolderName+DataName)        // Preparation
    For (i=0;i<Dimsize(Data,1);i+=1)        // Loop
        Killwindow  $(DataName+"_"+Num2str(i))
    EndFor
End


Be advised not to display the fit windows twice. Killing only works on the first instance. To fix this, you would need some checkup during the displaying routine that skips already displayed fits.

Happy fitting,
HJ

December 13, 2017 at 01:47 am - Permalink

Thank you HJ for step by step illustrations on how the code is developed towards the final step. My fitting is complete now. Igor Exchange forum has always been a great site to learn and develop skills for users like me.
One thing I would like to know is about unequal spaced y wave scaling. As Johnweeks mentioned, I created a y wave scaling (time) called "ywave" to match time for cols in 2D data, "Data". Then I run the following command: "Display; AppendImage Data vs {*,ywave}". Putting cursor in the image now displays the actual time and data however y scale in the displayed image does not match. Where am I doing wrong?
Thank you.
Tika

December 16, 2017 at 05:45 pm - Permalink

Hard to tell without details. Could share the graph (File -> Save Graph Copy) or the experiment ?
You can also send it by private message to me directly. (In the forum, click on my name and go to contact)

HJ

December 17, 2017 at 12:40 am - Permalink

Tika wrote:
Thank you HJ for step by step illustrations on how the code is developed towards the final step. My fitting is complete now. Igor Exchange forum has always been a great site to learn and develop skills for users like me.
One thing I would like to know is about unequal spaced y wave scaling. As Johnweeks mentioned, I created a y wave scaling (time) called "ywave" to match time for cols in 2D data, "Data". Then I run the following command: "Display; AppendImage Data vs {*,ywave}". Putting cursor in the image now displays the actual time and data however y scale in the displayed image does not match. Where am I doing wrong?
Thank you.
Tika

You will recall that my original post about this topic said:
Quote:
Most likely you just need to use a wave with one time value for each column of the wave. That would appropriate for curve fits, for instance.

The exception to this statement, that I hinted at but didn't explain, is when you use an auxiliary wave to display an image. For an image, you need the time wave to contain NCOLS+1 rows. This is because each "pixel" in an image display needs a position for the left *and* right edges; the final right edge of the last column requires one extra element.

Please read about this in our help: DisplayHelpTopic "Image X and Y Coordinates"

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

December 18, 2017 at 10:50 am - Permalink

Thank you Johnweeks and HJ, I will go through the help topic once again. Just for the info on how scaling looks like, I have attached the file.
Thanks,
Tika
Trial_2.pxp (1.63 MB)

December 18, 2017 at 11:36 am - Permalink

As John already pointed out, then number of data points matters and differs between images and waterfall plots. As far as I see, this applies also to the submitted example.
Happy Programming,
HJ

December 18, 2017 at 04:47 pm - Permalink