MathGroup Archive 1993

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

Search the Archive

RE: curve-fitting

  • To: mathgroup at yoda.physics.unc.edu
  • Subject: RE: curve-fitting
  • From: fulling at sarastro.math.tamu.edu (Stephen A. Fulling)
  • Date: Tue, 2 Mar 93 19:25:27 CST

>I am a Physics student at Fairfield University working on a
>project now that requires a curve fit with the following form:
>I = -I1*Exp[-t/T1] + I2*Exp[-t/T2] ...
>I would appreciate any response or input.
>                                Dan Kefford

This is the answer you will like least, but according to Forman S.
Acton, _Numerical Methods that Work_, this curve-fitting problem is
notoriously ill-conditioned, and should not be attempted by ANY method
other than experimental determination of the constants T1 and T2.
(This is on pp. 252-253 of the first edition, in the "Interlude:  What
NOT to Compute".)

I'm not a numerical analyst myself, so maybe Dan shouldn't just take my
word for it.  Would anybody else care to comment?

				      S. A. Fulling
				      Professor of Math,
				      Texas A&M Univ.






  • Prev by Date: Real vs. Integer paradox
  • Next by Date: Mathematica Bibliography
  • Previous by thread: RE: Real vs. Integer paradox
  • Next by thread: Mathematica Bibliography