Biexponential fitting (non-linear regression)on Matlab

Hi,

I have a problem whereby I need to fit a biexponential model to some data. I'm quite new to Matlab and the code is currently beyond my skill set.

The equations I need to complete are below, any help would be greatly appreciated. The full reference of the paper is: Maciejewski, H., Bourdin, M., Lacour, J. R., Denis, C., Moyen, B., & Messonnier, L. (2013). Lactate accumulation in response to supramaximal exercise in rowers. Scandinavian journal of medicine & science in sports, 23(5), 585-592.

Regards,

Ashley

[Rik]

Just a hint: try to reduce the content of the paper to the bare minimum, otherwise people may give this link, instead of ignoring the unnecessary details.

Answer 1

[Rik]

So you have a list of (La,t) combinations? With fittype you can tell Matlab what function it should fit to.

La0=La(t==0);
ft = fittype(sprintf('%e+A1*(1-exp(-gamma1*t))+A2*(1-exp(-gamma2*t))',La0),...
  'independent','t','dependent', 'La' );
opts = fitoptions( 'Method', 'NonlinearLeastSquares' );
opts.Display = 'Off';
opts.Lower = [0 0 0 0];%lower bounds (A1, gamma1, A2, gamma2)
opts.StartPoint = [1 1 1 1];
[fitresult, gof] = fit( t, La, ft, opts );

PS I expect you will need the curve fitting toolbox to use these functions

PPS I expect there is a better way to tell Matlab to treat La0 as a value from the workspace, but this will work. Maybe generating an anonymous function would be more elegant.

[Ashley Richardson]

Thanks for your help Rik,

Yes we have a list of (La,t) combinations. I don't have the curve fitting toolbox currently, but will work through your solution and try on the free trial of the toolbox (before purchase)if it doesn't work on the standard licence.