Services & Resources / Wolfram Forums
-----
 /
MathGroup Archive
2004
*January
*February
*March
*April
*May
*June
*July
*August
*September
*October
*November
*December
*Archive Index
*Ask about this page
*Print this page
*Give us feedback
*Sign up for the Wolfram Insider

MathGroup Archive 2004

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: again: nonlinear fit...

  • To: mathgroup at smc.vnet.net
  • Subject: [mg52145] Re: again: nonlinear fit...
  • From: p-valko at tamu.edu (Peter Valko)
  • Date: Fri, 12 Nov 2004 02:13:59 -0500 (EST)
  • References: <cmvd75$sb8$1@smc.vnet.net>
  • Sender: owner-wri-mathgroup at wolfram.com

I think the problem is that you consider L as a parameter. 
You can fix it by considering your measured x values and selecting the
largest to be L (or you can select the next nice round number as L).

Then your model will have only two parameters (v and Dz) and one
variable (x).
If FindFit still does not work, please provide typical values for the
two parameters and a typical range of the x variable.

Regards
Peter
 

db at ict.fhg.de (julia) wrote in message news:<cmvd75$sb8$1 at smc.vnet.net>...
> Hello,
> 
> I've posted some days before and i didn't get an answer.
> I think my question was not detailed enough and therefor I'll
> try it again.
> My experimental data have to be fit to this model:
> lamkonv[x_, {v_, Dz_, L_}] := 
> 1/(2*v^2*x^3)*((Dz*x/Pi)^0.5*(L*Exp[-L^2/(4*Dz*x)] - (L + v*x)
> *Exp[-(L - 2*v*x)^2/(4*Dz*x)]) + 0.5*(L^2 + Dz*x)*
> (Erf[L/(2*(Dz*x)^0.5)] - Erf[(L - 2*v*x)/(2*(Dz*x)^0.5)]))
> 
> where Dz,v,L are the parameters.
> (It is a 2-dim. convective-diffusion model)
> 
> I've tried an own code, nonlinearFit, NonlinearRegress and other
> optimization
> algorithms (quasiNewton..) ad I couldn't work it out.
> 1.The fit runs in a physically wrong range or 
> 2.I get imaginary values for the parameters or
> 3. I get the "nrlnum" -message or
> 4. The fit does nothing
> 
> Especially the parameter Dz behaves very strange. The value of Dz i
> about 10^3 times smaller than the other parameters. I've read in a
> posting before, that this may cause problems.
> 
> Can anybody help me with that?


  • Prev by Date: Re: Fibonacci based sum that is b-normal on binary numbers
  • Next by Date: Re: Re: First question
  • Previous by thread: again: nonlinear fit...
  • Next by thread: minimal Pisot tile {2,3] type definition in Mathematic