I have an m x n matrix where every row can be fit with a1*exp(-k1*t)+a2*exp(-k2*t)+... 
exponentials, and I like to find the a, k set which gives the best fit 
for the hole surface. I was thinking to use lsqnonlin or lsqcurvefit but 
I cannot find the way how to use it for a problem like this as ydata 
should be a vector. Does someone have any idea?

"Andras " <andras.lukacs@gmail.com> wrote in message <hte2p6$n6g$1@fred.mathworks.com>...
> I have an m x n matrix where every row can be fit with a1*exp(-k1*t)+a2*exp(-k2*t)+... 
> exponentials, and I like to find the a, k set which gives the best fit 
> for the hole surface. I was thinking to use lsqnonlin or lsqcurvefit but 
> I cannot find the way how to use it for a problem like this as ydata 
> should be a vector. Does someone have any idea?

Hi,

I'm looking for the same problem. A kinetic dataset that has to be fitted with a multiexponential function. The amplitudes for every curve can vary but the time constants k should be the same for every curve. The whole problem should be worked out as a global fit, is this possible?

My dataset:

xdata -> m x 1 time
ydata -> m x n absorbance
zdata -> 1 x p wavenumber

z can be ignored because it has nothing to do with the kinetics, but how to do a multiexponential global fit with matlab for a problem like this. Thanks in advance for every hint.

Greetings

Fred 

"Frederik Grosserueschkamp" <fred@bph.rub.de> wrote in message <i4gk0e$7it$1@fred.mathworks.com>...
> "Andras " <andras.lukacs@gmail.com> wrote in message <hte2p6$n6g$1@fred.mathworks.com>...
> > I have an m x n matrix where every row can be fit with a1*exp(-k1*t)+a2*exp(-k2*t)+... 
> > exponentials, and I like to find the a, k set which gives the best fit 
> > for the hole surface. I was thinking to use lsqnonlin or lsqcurvefit but 
> > I cannot find the way how to use it for a problem like this as ydata 
> > should be a vector. Does someone have any idea?
> 
> Hi,
> 
> I'm looking for the same problem. A kinetic dataset that has to be fitted with a multiexponential function. The amplitudes for every curve can vary but the time constants k should be the same for every curve. The whole problem should be worked out as a global fit, is this possible?
> 
> My dataset:
> 
> xdata -> m x 1 time
> ydata -> m x n absorbance
> zdata -> 1 x p wavenumber
> 
> z can be ignored because it has nothing to do with the kinetics, but how to do a multiexponential global fit with matlab for a problem like this. Thanks in advance for every hint.
> 
> Greetings
> 
> Fred 

Did you try use the surface fitting tool and define custom equation?

> Did you try use the surface fitting tool and define custom equation?

In think that it is only possible to load vectors into the surface fitting tool but not the matrix ydata with the spectral information. But if there's a way I'll be very happy.