EzyFit 2.40

A free curve fitting toolbox for Matlab

Description

The Ezyfit toolbox for Matlab enables you to perform simple curve fitting of one-dimensional data using arbitrary (non linear) fitting functions.

EzyFit adds a new menu to your figure windows, which allows you to easily fit your data with predefined or user-defined fit equations, including selection of your data (outliers removing) using the ``Data Brushing'' tool (available since Matlab 7.6).

This toolbox also provides a set of command-line functions to perform curve fitting 'programmatically': you just have to type something like showfit('c+a/x^n') and EzyFit gives you the values for c, a and n and shows you the curve!

Visit http://www.fast.u-psud.fr/ezyfit/ for additional info.

Installation and system requirements

EzyFit needs Matlab 7.0 or higher. It has been tested under 7.0 to 7.10 (R2010a), but mainly under Windows XP. The command-line functions (e.g. ezfit, showfit...) work equally well on all systems. However graphical operations (e.g. showslope, getslope...) may not be fully stable, especially on non-Windows systems, and/or for recent Matlab releases.

1. Download and unzip the EzyFit Toolbox in a directory somewhere in your system. For instance, in a Windows XP installation, the directory My Documents/MATLAB/toolbox/ezyfit may be a good location. Do NOT install the toolbox in the Matlab directory itself (Program Files/Matlab directory in Windows). If you upgrade from an older version, first empty the previous directory.

2. From the menu 'File > Set Path', click on 'Add Folder' (NOT 'with subfolders') and select the ezyfit directory. Click on 'Save' and 'Close'.

3. Restart Matlab. From the Start button, select Toolboxes > EzyFit to check that the toolbox is correctly recognized by your system.

4. If you want to always have the Ezyfit menu in your figures, type efmenu install. This will create or update your 'startup.m' file in the directory /toolbox/local of your Matlab installation.

Note: If you upgrade Matlab and you want to use your previous Ezyfit installation, you just have to follow the steps 2-4.

Examples:

plotsample power % a power law
showfit c*x^n

plotsample hist % an histogram
f = ezfit('gauss'); % fits with a Gaussian
f % displays the values of the coefficients

• MATLAB release: MATLAB 7.0.1 (R14SP1)

[Jack Chen]

Looks like a great free toolbox for matlab. However, I'm getting the following error:

>> efmenu install
??? Error using ==> load
Unable to read MAT file d:\matlab\easyfit\userfit.mat

[Frederic Moisy]

Dear Jack Chen,
This error does not occur with R14.
Some features of this toolbox (eg, nested functions) need R14. Sorry.

[John D'Errico]

While I did not try out the menu interface, the command line interface was indeed easy to use. A splendid tool.

[Blair Greenan]

Excellent toolbox for curve fitting. Any questions I had were quickly answered by the author.

[Dietrich Lueerssen]

I do like this toolbox, and it works well for me. It is really useful that there are both a GUI and a programmatical access, well done.

I do have a suggestion for future versions: at the moment, parameters of the fit such as its color are hardcoded into a file (fitparam.m). It would be helpful to be able to control that programmatically, particularly if you want to show multiple fit curves for the same data set (e.g., to show that one fit is better than another one).

[Frederic Moisy]

Dear D. Lueerssen, Thanks for your comment. The new version (2.10), submitted 30 oct 2006, now allows for specifying fit parameters programmatically. Example: showfit('a*x+b','fitcolor','red',...) - The author.

[Jonas Semoi]

 

[Ravi Prakash]

Hi,
Its superb toolbox you ahve provided to Matlab users. I was struggling to fit a smooth curve for amplitude analysis of seismic data. This toolbox does it in fantastic way.
Thanks
Ravi

[Andrew Isaacs]

Excellent work. Did what I needed (multiple power terms) when no-one else could - at least for free. Easy to install and work with, and great solutions.

[Matt Lopz]

Excellent Utilty.

[atanas atanasov]

 

[Josiah Haswell]

Fabulous

[MArta Castillo]

Excellent

[david abks]

Great - well done!!!!!!!!

[Andres Leal]

Great software! I was able to do curve fits using custom equations that wouldn't be fitted with KaleidaGraph.

[D J]

a nice tolbox...best alternative to matlab curve fitting toolbox which sadly doesnot come for student version of the software.

[Woody Schneider]

Great, but it behaves sketchily on the Mac OS X Intel version...

[Gilles Bertrand]

A great tool! Fill a gap in matlab functionalities. Good interface, good integration into matlab, merci Frederic M.!!

[Alex Liberzon]

Very useful.

[Chintan Trivedi]

 

[andy yang]

So splendid!

[Hauke Schramm]

 

[Mickael PARIS]

 

[Joe M]

 

[m la]

 

[Mark Dean]

Thank you for a useful and well writtten program!
I was able to write a small program to easily choose whether to fix or vary some of the fit parameters. It would be useful if this was included as standard.

[A Rose]

Perfect! Just what I needed to estimate parameters for over 500 chick growth curves.

[R B]

Very good and powerful toolbox

[P. Sci]

 

[sri murthy]

Great alternative to cftool...excellent

[K E]

Very helpful and intuitive.

[Jason Ruth]

Fantastic program, very straightforward and easy to use.

[Nirvikar Dashora]

Please tell me how to Unistall the ezyfit toolbox from MATLAB.

[N Davidson]

Simple to use and very useful.

[? ?]

Excellent

[P Jop]

very usefull with command-line functions

[M Lidierth]

Used the GUI menu without any hitches. Intuitive and easy-to-use. Excellent.

[Alex Liberzon]

great support and improvements over the last two years.

[Jim S]

Excellent and smooth!

[Chris Calvert]

Just downloaded this, it looks great! Any functions I have tried (so far) work very easily and smoothly. Are there any plans to include lorentzian fitting in the tool box?

[Janina Madoff]

impossible to remove their menu from your gui figures and uninstall cleanly

[Thierry Dalon]

Janina and Nirvikar: do not put bad ratings only because you don't know how to use Matlab. At least read the documentation before, contact the author before or ask for support in the Newsgroup. Your rating is not fair!
To remove the menu from figures use>> efmenu off. To remove the Toolbox simply remove from Matlab path or entry in startup file if you used efmenu install.

To the author : to avoid function naming conflicts, it may be better to name all the functions in the Toolbox with a common prefix like 'ef'.

[Timo]

Take a look at the known problems: "If the Ezyfit toolbox has been installed using efmenu, all GUI figure windows created using GUIDE have the Ezyfit menu included. Uninstalling the Ezyfit toolbox produces an error, since the GUI looks for the menu which is not available. No solution for the moment."
Using efmenu off does not remove the menus.
This is a clear minus for an otherwise very useful toolbox.

[Frederic Moisy]

 

[Frederic Moisy]

(Sorry for the unintentional rating of my own file on my previous post!)

TO UNINSTALL THE EZYFIT TOOLBOX:
1. Edit the 'startup.m' file (type 'edit startup') and remove the 2 lines starting with 'efmenu' and 'fprintf()'
2. Go the menu 'File', 'Set Path', and remove the directory 'ezyfit'
3. Restart Matlab

CONCERNING THE ISSUE OF THE EZYFIT MENU IN SAVED FIGURES AND GUIs:
Sorry for this bug, I have no idea on how to fix it. Suggestions are welcome. Note that you can use the Ezyfit toolbox without using the Ezyfit menu. For this, apply step 1 above.

Hope this helps - The author.

[Janina]

Thierry Dalon -
No need to sling insults like a common blogger. I had assumed better of Matlab users. I am referring to the bug the Frederic Moisy himself identified - that it is difficult to remove the ezfit menu from previously saved figures. I have indeed removed the ezfit menu. I reloaded my figure generating scripts, and all was well. Thank you Frederic, for your feedback.

[Soumitra Deshmukh]

Does it work with earlier versions of Matlab such as 6.5 ?

[Michael Jordan]

 

[Cezar]

Excellent tool! Many thanks!!!

[Thanos]

Very useful and well explained. Thank you!

[Roy]

 

[Sonali]

 

[Florian]

 

[Clemens Listner]

Thank you for this great toolbox

[Erdal Bizkevelci]

 

[Farah]

I just downloaded this and it's great! Very easy to use and non-cluttered. 5 over 5! Is there any where I can obtain more in-depth information about how it works and what the output values really mean? That would be superb.

[Antony Lee]

The package is really great! However, I would like to know whether it is possible to have some fixed parameter values when using ezfit, eg.
ezfit(x,y,'y(x)=a*x+b')
assuming that I know the actual value of a and want only the fit for b. (Of course if a=3 or some other value I could write ezfit(x,y,'y=3*x+b'), but I want to include this in a script, so a would be a variable...)

[Roy]

 

[Arnold]

Awesome toolbox. Instrumental in my research.

[Erdal Bizkevelci]

 

[Andrew Barrow]

Fantastic tool. This is incredibly useful. Especially if you use the "data paintbrush" tool in the figure window.

[Stefan Gachter]

Excellent tool!

[Muhammad Tauha Ali]

splendid tool. inclusion of statistical tests such as RMSE, t-statistics and R-squared statistics in ezyfit menu will enhance the functionality of the toolbox.

[Dave Robinson]

Nice application, seems to do everything it says on the box

[Andres Klever]

Hi Frederic, well im sending you this mail because i have a problem with this tool, first of all say to you thanks for this tool i think that is really good, ok , so there are 2 problems:

1) i have a problem with the weight
 In order to perform a weighted fit on this data, the vectors y and dy have to be merged into a 2-by-N matrix and given as the second input argument to ezfit. Compare the results for the usual and weighted fits:
fw = ezfit(x, [y;dy], 'exp');
showfit(fw,'fitcolor','red');
f = ezfit(x, y, 'exp');
showfit(f,'fitcolor','blue');
my problem is that when i do this (exactly the same) the data [y;dy] is from a different size than x, so obiously i get this error:

??? Error using ==> ezfit at 272
X and Y dimensions must agree.

if you can help me with this it would be great..

2)the other problem is that when i plot more than one fit for the same data, the plot just shows one legend that is pasted into the other legend, hahaha, so i just can see one of the 2 legends that should appear (showing me the different equations)...

ok, frederic sorry for asking you this questions but i really need to know this,

thanks very much,

[Frederic Moisy]

Dear Andres,

Problem #1: This occurs if y and dy are column vectors instead of row vectors. If this is the case, then just replace [y; dy] by [y, dy]

Problem #2: Move the equation box using the mouse.

-- The author

[ihsan]

Great toolbox!!! Many Thanks.

[Massimo]

Dear Frederic,
is it possible to associate the error to the estimated parameters?
Thanks

[Ritesh]

 

[Arthur]

Very usefull!

[Mathieu]

This tool is excellent! It makes life so much easier!
Thanks Fred!!!!
And BRAVO!!!!

[Oliver]

 

[anurag pandey]

it's a fabulous tool bar, which solved my problem i was struggling to get done for quite some time. I have a query about it, that what you do if you want to fit just a part of your data(curve).

[Melanie Mu]

Great tool! It's almost perfect! I am only missing two things; I would really love if it could also:
1) fit data with error bars in both directions
2) display the errors associated with the fit parameter results.

[Xiao Ling]

very useful tool! Thanks a lot

[Shahriar]

Excellent toolbox! Very useful. Thanks!!

[Brad Ridder]

Its amazing in every way.
A gold-standard for user-submitted toolboxes.

A job well-done! Bravo!

[Matthias]

You're a champion for creating this!
Thanks very much...

I do have a couple of questions I would like to ask:

When you say "There is no way for the moment to change this number of iterations in EzyFit", I am a bit confused by this. In fminsearch the default for 'MaxInter' and 'MaxFunEvals' are both '200*numberOfVariables'.

I have set these to 2000000*numberOfVariables but I still run into the exact same error when using:
ezfit(x,y, a1*exp(-b1*x)+a2*exp(-b2*x)+y0)

The error is:
Exiting: Maximum number of function evaluations has been exceeded
         - increase MaxFunEvals option.
         Current function value: 30502.906917

Also, since MaxInter and MaxfunEvals both must be integer values, I do not understand how the iteration above can have decimal values. Or maybe I am misinterpreting it?

The other thing I wanted to say is that if run fminsearch and call fitfun as per the code below, fminsearch is able to generate a perfect 3 exponent decay function in 175 iterations (i.e., much lower than 30502.906917). I should think ezfit should be able to achieve the exact same result if allowed to iterate within definable constraints (see TollX and TollFun below), because it also uses fminsearch.

Finally, probably a dumb question, how come 'showfit' will work when I supply a 3 (or 2) exponent function but ezfit fails with the above iteration error?

Looking forward to reading your answers......this is otherwise a brilliant toolbox.

Best regards,
Matthias

code:
-------
% Define initial paramters

t = x;
y = y;
    % Grab x and y values generated in ginput

TolFun = .000000001;
TolX = 0.1;
    % TolFun = Termination tolerance on the function value
    % TolX = Termination tolerance on x

% Plot x and y

set(0,'DefaultFigureWindowStyle','docked')
    figure;plot(t,y,'ro')
    hold on
    h = plot(t,y,'b');
    hold off
    title('Input data')
start = [0;0;0];
    % Define starting lambdas (initial guesses)

outputFcn = @(x,optimvalues,state) fitoutputfun(x,optimvalues,state,t,y,h);
    % User defined output function for each iteration of fminsearch

options = optimset('OutputFcn',outputFcn,'TolX',TolX,'TolFun',TolFun, 'Display', 'iter');
    % Options for fminsearch

estimated_lambda = fminsearch(@(x)fitfun_v2(x,t,y),start,options);

[Christopher Coello]

Great, great toolbox, really useful.
One possible improvment: to handle subplot command in the display using showfit. Now, when doing subplot, the info about the fit always shows on the first subplot (subplot(xxx,xxx,1)).

[Siva]

Just great !

[Patrick]

Very helpful tool! In stock form, can't handle completely arbitrary fit functions; EQ2ML doesn't convert parameters in expressions such as 'myfitfun(K,x)' to the required m(i) format for fminsearch, since it treats 'K,x' as a single token. This is quickly fixed by adding ',' to the list of delimiters used by FINDWORD, after which it handles that input just fine.

Along those lines, it would be nice if you could specify the fit function using a handle; restricting the 'x' variable to be either the first or last free variable in the handle/anonymous function should allow that.

[Liang Jin]

Great features and easy to use.

[Basim Kakakhel]

An excellent utility, I am having little trouble in running the following.

I have two vectors of 360 elements each.

First vector is y which contains the pixel number from 1-360

the second vector is c, it contains the values for the y vector.

the aim is to fit a function similar to this

y=a(1-exp(b*c))

I am more interested in the a coefficient. However i am getting the following error.

 ezfit(y,c,'y=a*(1-exp(-b*c)')
??? Error using ==> ezfit at 549
Fit: error during the fminsearch procedure

[Frederic Moisy]

Dear Basim Kakakhel,
Please try the following: ezfit(c,y,'y(c)=a*(1-exp(-b*c)')

[Al Ka]

Very good plugin but I have a serious problem with evalfit(). I have the following code:

x = [1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749];
x= x';

y = 1.0e+03 * [1.0408 1.0392 1.0376 1.0362 1.0348 1.0335 1.0323 1.0312 1.0302 1.0293 1.0285 1.0278 1.0271 1.0266 1.0261 1.0258 1.0255];
y=y';

f = ezfit(x,y,'y=20-a*10*log(x)');

figure; plot(f.x,f.y, 'b-');
figure; plot(x,evalfit(f,x),'r-');

If you run it you will see that the two plots are way off while they should be almost identical. I guess. Please let me know what I am doing wrong.

I want to use evalfit to extrapolate some data. Thank you in advance!

[Wesley]

 

[Roman Voronov]

Please add support for NaN values

[Kassi]

I would also like support for NaN values, please! :)

[Francois Anquez]

Very Nice tool !
Easy and rapid handle. Something that sometime miss me with matlab...

Thanks to the author.

[rene]

 

[rene]

Brilliant plugin, though i had problems getting it to work with matlab 2010a.
If anybody else is having problems with getslope.m you have to change line 76 from
if v>=7.4
to
if v>=7.4 || regexp(strv,'.*R20\d\d.*')==1

This should correct the version detection.

[vick Audi]

Problem with unistallation:

Even after following your following recommended steps (under full unistallation):
Full uninstallation

In order to remove the Ezyfit toolbox from your Matlab installation, follow the 3 steps:

1. Edit the 'startup file' by typing

edit startup
        
and remove the following lines

% These lines have been added by 'efmenu install' (01-Apr-2008 23:10:11):
efmenu; % Includes the EzyFit menu for all new figure.
fprintf(' To get started with the EzyFit toolbox, select EzyFit from the Help browser.\n\n');
        
2. From the menu 'File > Set Path', select the directory '../ezyfit', and click on 'Remove' and 'Save'.

3. Restart Matab.

%*************************************%
I'm getting the following error when I run my GUI:
%*************************************%

??? Error using ==> struct2handle
Undefined function or variable 'efmenu'.
 
??? Error using ==> struct2handle
Error while evaluating figure CreateFcn

%*************************************%

This is getting a little annoying now. Kindly suggest a fix. Thanks

[Carl]

 

[Thorsten Maly]

 

[Homerarg]

 

[João Faria]

 

[Nathan Tomlin]

 

[Yvonne Haesevoets]

Great tool. Thanks for sharing.