How can I extract X and Y from gauss2 fit or any other fit (poly2, poly 3, ...)?

Hello,

I have a txt data and I want to fit this data in other to calculate de beam radius of this gaussian plot but I dont know how can i have the X and Y of the fit. My code is:

  % scan the column until end1 and saves the Y and E     
  for h = end1(loop):-1:lastpoints1
   for f = 1:numfiles
        E2 = E(:,f);
        Y2 = Y(:,f);
        gaussfit = fit(Y2,E2,'gauss2'); 
        
        %I need to have the new Y and X (or E) to use the condition "if" below:
        
     % if NEW_E(h,:) >= m(f)
     %     x(h,f) = NEW_Y(h,f);
     %     y(h,f) = NEW_E(h,f);
     % end 
    end 
  end

You can see that everything is inside a loop. This is because I have more than one file to do the same thing: extract the X and Y of each file, fit, extract the new X and Y and scan until the condition "if" is true and save the value X and Y when the condition "if" is true.

I would be very grateful if someone knew how to extract the X and Y from the fit.

I tested the fit and it is enough for what I need but I don't know how to get the curve in an array or matrix.

Many thanks.

[Stephen23]

https://web.archive.org/web/20210207191956/https://www.mathworks.com/matlabcentral/answers/580863-how-can-i-extract-x-and-y-from-gauss2-fit-or-any-other-fit-poly2-poly-3

Answer 1

[Adam Danz]

The fit function merely fits your pre-existing (x,y) values and returns a fitobject that contains parameter values you can use to plot the function. There are no other x or y values other than the ones you provided to the fit function.

You can create any set of x values and compute the y values from the fitobject.

Or, you can plot the fitobject and let Matlab create and x and y values of the fitted line and then extract them.

Here's how to do both.

Option 1: Create any set of X values, then compute Y

Pros

• Flexibility; you can choose any range of X values and at any interval.
• No need to plot the the data if you don't need to.
• You'll learn more using this method.

% Load built-in dataset
load census; 

% Fit quadratic polynomial
f=fit(cdate,pop,'poly2'); 

% Plot the raw data (if you want to)
clf()
plot(cdate,pop, '.')

% Create x values using the min & max of the existing data
X = linspace(min(cdate),max(cdate),100); 

% Compute the Y values.  The function to use is defined by the fitobject output. 
% f = 
%      Linear model Poly2:
%      f(x) = p1*x^2 + p2*x + p3     % <----------------- HERE
%      Coefficients (with 95% confidence bounds):
%        p1 =    0.006541  (0.006124, 0.006958)
%        p2 =      -23.51  (-25.09, -21.93)
%        p3 =   2.113e+04  (1.964e+04, 2.262e+04)
%
% Or you can look it up here
% https://www.mathworks.com/help/curvefit/list-of-library-models-for-curve-and-surface-fitting.html#btbcvnl
% Or you can get it from the formula(f) command.
coefs = coeffvalues(f); 
Y = coefs(1)*X.^2 + coefs(2)*X + coefs(3);  

% Plot Curve (if you want to)
hold on
plot(X,Y, 'r-')

Option 2: Plot the fit, then grab the (X,Y) values Matlab chose

Pros

• Matlab does the heavy lifting.
• Less room for user error.

% Load built-in dataset
load census; 

% Fit quadratic polynomial
f=fit(cdate,pop,'poly2'); 

% Plot the raw data & fit
h = plot(f,cdate,pop); 

% Get the handle to the fitted curve
curveHandle = findobj(h,'DisplayName', 'fitted curve'); 

% Get the X and Y values Matlab chose
X = curveHandle.XData; 
Y = curveHandle.YData;

[Tay]

Hi Adam !!

Yes, that's what I need.

I'll try option one, but option two has already solved my problem.

Thanks a lot for the help !! Really helped me.

Cheers.

[Adam Danz]

Glad I could help!

[Adam Danz]

indexMax(loop) = find(Y1(:,loop) == max(Y1(:,loop)));

can be replaced by

[~, indexMax(loop)] = max(Y1(:,loop));

if you expect only 1 max value.

I'm lost here: "...but some of those values are bigger than the others..."

What values? What other values?

Also,

   for h = 1:end1(loop)

are you sure the (loop) should be there?

[Tay]

Hi Adam,

Thanks for your reply.

"What values? What other values?"

Sorry I wasn't clear.

I'm talking about the "end1" vector. As you can see the end1 vector which contains the indexes of the maximum values of Y1 has different values: 967 876 802 716 629 566 516 464 403. That is, the indexes of the maximum values of each column are different.

I want to scan each column to its maximum value. For this I had to find the maximum value index for each column (I used the "find" for this and saved it in the vector end1).

To access these indexes and be able to scan I put the for that way:

   for h = 1:end1(loop)

This loop goes 1 to numfiles that is 1:9 (first line in my code).

But at first it is not working because the matrix "x" and y "y" that should have 9 columns with different number of rows always has the same number of rows (967 - the biggest number).

Is it possible for me to have a matrix wiht nine columns and different numbers of rows?

[Adam Danz]

Both x and y are currently defined by (h,f) unless you're preallocating those variable (which I don't see in the code you shared).

If I understood the goal better I might be able to give better adivce but I don't understand what is suppos to be happening here. Perhaps you could explain it in a few sentences.

[Tay]

I'm sorry for the long text. I tried to explain it correctly. I hope you understand..

Thank you very much in advance !!

[Adam Danz]

Those variable names make it really confusing to follow.

It sounds like you're trying to find the two x-coordinates where the curve intersects with the line at y=ymax*0.37. Of course you could solve for y, knowing the function and the x-inputs.

Alternatively, you could very closely approximate the intersections using the intersections function from the file exchange.

 [x0,y0] = intersections(x,y, [min(xlim),max(xlim)], ymax*0.37*[1,1]);

where x and y are the coordinates used to plot the curve and xlim is Matlab's xlim() function.

Then, plot the results for confirmation

 hold on
 plot(x0,y0, 'm*')
 yline(ymax*0.37)