Parameter fitting using nlinfit (high order ODEs)

I am trying to fit data on multi-dimensional ODEs:

function output = myfunc(pars,tspan)
      % we just rename the variable here for use in the ODE
      k = pars;
          function dCadt = ode(t,Ca)
              % this is the ODE we are fitting to
              dCadt(1) = -k(1)*Ca(1);
              dCadt(2) = k(2)*Ca(1)*Ca(2);
          end
      % Initial concs. of Ca.
      Cao = [2 3]';
      [t,Ca] = ode15s(@ode,tspan,Cao);
      output = Ca;
  end % myfunc

Calling function:

    tspan = [0 0.1 0.2 0.4 0.8 1]';
    Cadata = [2.0081  1.5512  1.1903  0.7160  0.2562  0.1495;
              2.0081  1.5512  1.1903  0.7160  0.2562  0.1495]';

    parguess = [1.3 2]; % nonlinear fits need initial guesses
    [pars, resid, J] = nlinfit(tspan,Cadata,@myfunc,parguess)

However it always give this error:

Error using nlinfit (line 185)
Requires a vector second input argument.

When I run same program with lsqcurvefit, it runs smoothly!

Please anyone help me to sort this error!

Answer 1

[Star Strider]

That has to do with differences between lsqcurvefit (that will fit matrix dependent variables) and nlinfit that will only fit vector dependent variables.

[Learner]

Thanks for the clarification.

[Star Strider]

My pleasure!

[Learner]

Is it possible to use 'nlparci' with lsqcurvefit?

Or do you know any other command similar to nlparci which can be used with lsqcurvefit?

Thanks in advance.

[Star Strider]

I don’t remember if nlparci will work when you’re fitting two dependent variables, since it’s been a while since I tried that. I believe it will, but I know nlpredci will not.