Polyfit works in command window, but returns incorrect values in M-File

I can't go into too much detail on the application of my code because it is for a confidential project. What I am working on though interfaces with Prolith, a microlithography simulation tool. So posting the entirety of my code here would be useless. My process is as follows:

1. Run a bunch of simulations in Prolith from MATLAB
2. Get data from Prolith from simulations
3. Analyze data to determine response
4. Use polyfit function to model response quadratically

My function returns four values:

• dCDA: One of the responses
• sCDA: The second response
• dCDA_model: the modeled dCDA
• sCDA_model: the modeled sCDA

Now here is my problem, when I run my function it returns the incorrect values. However, running the same code from the command window returns the correct values. My x values are in a vector called ZMag.

EDIT: See comment below for corrected data

 ZMag =
  0 1.93 3.86 5.79 7.72 9.65

 dCDA =

   -0.0811    0.1554    0.4109    0.7142    1.0336    1.3741      
    0.0635    0.4425    0.8215    1.2006    1.5791    1.9525    
   -0.0099   -0.0079   -0.0078   -0.0090   -0.0117   -0.0159   
    0.0526   -0.1442   -0.3442   -0.5637   -0.7802   -0.9938   

sCDA =

   -0.1050   -0.5570   -1.0011   -1.4324   -1.8442   -2.2308   
   -0.1034    0.2494    0.6041    0.9594    1.3145    1.6516  
   -0.0187    0.1713    0.3622    0.5530    0.7437    0.9392  
   -0.0908   -0.2982   -0.4927   -0.6840   -0.8622   -1.0330  

 dCDA_model (incorrect)

    151.5506  -15.3017  -19.9054    5.5466
    21.1869   38.5986    0.9280  -21.4042
    -0.0715    0.0594   -0.0144    0.0662

 sCDA_model (incorrect)

    155.0786  -33.6812  -19.3941    4.9035
    -50.9605   36.9912   20.3325  -18.6355
    -0.0700   -0.1135   -0.0295   -0.1105

 dCDA_model (correct)

    0.0041   -0.0004   -0.0005    0.0001
    0.1098    0.2000    0.0048   -0.1109
   -0.0715    0.0594   -0.0144    0.0662

 sCDA_model (correct)

    0.0042   -0.0009   -0.0005    0.0001
   -0.2640    0.1917    0.1053   -0.0966
   -0.0700   -0.1135   -0.0295   -0.1105

And finally, the code that generated these anomalies:

for i = 1:n
  dCDA_model(:,i) = polyfit(ZMag,dCDA(i,:),2);
  sCDA_model(:,i) = polyfit(ZMag,sCDA(i,:),2);
end

Any ideas as to what could be causing this disparity in results between the command window and M-File? I am using MATLAB R2007a.

Thank you.

[Zac]

I truncated my matrices because they are rather long, I didn't even think it would change the modeled curve. Anyway, lets just work with one complete vector.

ZMag = 0 1.93 3.86 5.79 7.72 9.65 11.58 13.51 15.44 17.37 19.3

dCDA = -0.0811 0.1554 0.4109 0.7142 1.0336 1.3741 1.7407 2.1389 2.5751 3.0548 3.5848

sCDA = -0.1050 -0.5570 -1.0011 -1.4324 -1.8442 -2.2308 -2.5865 -2.9047 -3.1793 -3.4040 -3.5827

Answer 1

[Sean de Wolski]

EDIT

Both look good to me...

dCDA_model = polyfit(ZMag,dCDA,2);  %your new data
sCDA_model = polyfit(ZMag,sCDA,2);

subplot(121)
plot(ZMag,dCDA,'b-',ZMag,polyval(dCDA_model,ZMag),'rp')
subplot(122)
plot(ZMag,sCDA,'b-',ZMag,polyval(sCDA_model,ZMag),'rp')

And they both work in mfile/desktop. That's why I really am wondering if you're calling ML's polyfit

which polyfit

?

[Zac]

I truncated my matrices because they are rather long, I didn't even think it would change the modeled curve. Anyway, lets just work with one complete vector.

ZMag = 0 1.93 3.86 5.79 7.72 9.65 11.58 13.51 15.44 17.37 19.3

dCDA = -0.0811 0.1554 0.4109 0.7142 1.0336 1.3741 1.7407 2.1389 2.5751 3.0548 3.5848

sCDA = -0.1050 -0.5570 -1.0011 -1.4324 -1.8442 -2.2308 -2.5865 -2.9047 -3.1793 -3.4040 -3.5827

Regardless of getting the same results, doesn't it seem strange that it's fitting such a large parabola (in the incorrect case) to such small numbers?

[Zac]

Also, prior to calling the polyfit function there is the end of the for loop in which *CDA are calculated. That is all that is prior to the call.

Answer 2

[Zac]

Update: Tried putting code into its own function in a separate m-file and it returns the same value. How could I be calling MATLAB's polyfit from the desktop, but some other polyfit within the function? Doing a search for "poly" through my code finds only the spot in question. Could you guide me in finding out if I am really not calling MATLAB's function?

Answer 3

[Zac]

It is calling MATLAB's polyfit,

which polyfit
C:\Program Files\MATLAB\R2007a\toolbox\matlab\polyfun\polyfit.m

I just made a callback script to call polyfit outside of the function. It's clunky, but it works. I'm not sure why though.