what toolbox do I need to fit an exponential with bounded parameters?

Hi,

I want to fit an exponential function to my data, and force this exponential to be positive and decreasing (i.e. y=a*exp(b*x) with a>0 and b<0).

I figured this is what I need (built using examples from helfile):

fo = fitoptions('Method','LinearLeastSquares',...
               'Lower',[0,-Inf],...
               'Upper',[Inf,0]);           
ft = fittype('a*exp(b*x)','independent','x','coefficients',{'a','b'},'options','fo');
coeffs=fit(data(:,1),data(:,2), 'ft');    

However, I get a weird error message while compiling:

To use 'copy', at least one of the following products must be licensed, installed, and enabled:
  RF Toolbox
  Simulink
  Simulink Control Design
  Simulink Design Optimization
  Stateflow

Error in fittype>iParseParameters (line 567)
            obj.fFitoptions = copy( value );    
            .....

Do I really need additional toolboxes, or is there a wrong syntax in my code that I can't see?

Thanks a lot in advance,

Francois

[Torsten]

You need the curve fitting toolbox.

[Francois]

I have it, and it is activated too (otherwise Matlab would not even recognize the fit and fitoption functions)

[Torsten]

Why do you write under "fitoptions" that it's a linear least squares problem ? It is not.

Further use

coeffs=fit(data(:,1),data(:,2), ft);

instead of

coeffs=fit(data(:,1),data(:,2), 'ft');

Answer 1

[Francois]

Ho thanks! Sorry I am new to Matlab and tend to get confused over what should be a string and what should not.

I finally managed now, actually skipping the whole fittype call and putting the fitoption as an argument to the fit function directly. I do not think this construction is very well explained in the help file...

for the record, this is how I made it work:

options = fitoptions('exp1');
options.Lower = [0 -Inf];
options.Upper = [Inf 0]; 

coeffs=fit(data(:,1),data(:,2), 'exp1',options); 

Thanks again for your help, Torsten.

Answer 2

[Image Analyst]

You do not need Simulink or the Curve Fitting Toolbox. You can use the more common Statistics and Machine Learning Toolbox and use fitnlm (Fit Non Linear Model). I'm attaching a demo.

Answer 3

[Francois]

And this is probably the toolbox I will actually need. Thanks for the tip..

[Image Analyst]

Not sure who your "Answer" here is talking to...

Or what toolbox you're referring to.

[Francois]

I was actually trying to reply to you... Statistics and machine learning Toolbox looks like what i should have requested from my office, had I known that it did include a fitting tool.