Fitting numerical data
- To: mathgroup at smc.vnet.net
- Subject: [mg62022] Fitting numerical data
- From: Tomek <rtomek at hotmail.com>
- Date: Wed, 9 Nov 2005 03:45:48 -0500 (EST)
- Sender: owner-wri-mathgroup at wolfram.com
I am trying to find a best fit of my data to a function, which is a solution of a partial differential equation integrated over one of the variables:
SOLN = Function[{p, avg},
sol = NDSolve[D[u[r,t],t]... <diff equation>];
NIntegrate[Evaluate[First[u[r,t] /. sol]], {r, rmin, rmax}]/. t-> p]
When I compute the values at certain points for SOLN, everything works fine, but when I try to use FindFit to actually find the fit of the data:
FindFit[data, a*SOLN[t, av], {a, av}, t]
it fails with 'Unable to solve for the fit paramters; the design matrix is non-rectangular, non-numberical or could not be inverted.'.
The data provided allows for unique solution of the a,av paramters. Is it possible to solve those type of functions with FindFit. How to go around doing that?
Tomek