Mathematica 9 is now available
Services & Resources / Wolfram Forums / MathGroup Archive
-----

MathGroup Archive 2012

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

Search the Archive

Non-trivial Non- linear fit

  • To: mathgroup at smc.vnet.net
  • Subject: [mg126596] Non-trivial Non- linear fit
  • From: Meriens <meriens at gmail.com>
  • Date: Tue, 22 May 2012 05:20:33 -0400 (EDT)
  • Delivered-to: l-mathgroup@mail-archive0.wolfram.com

Hi, I've got a problem with a nonlinear fit.
My function is this:
B1= a1+b1/tr+c1/tr3
B2= a2+b2/tr+c2/tr3
B3=a3+b3/tr+c3/tr3
Btot=x*B1+y*B2+z*B3
I've got two dataset (some values are given below), one containing the Btot and Tr ed one containing x,y,z. My target is to make a regression to obtain the values a,b,c of every equation using only one regression.
In other words,
If I make a single regression for every x,y,z of the second dataset, my regressione will be indipendent and the coefficients will appear different for every regression.
So I want find a way to make only one regression by changing into this operation Btot, Tr, x,y,z. So my coefficients will not be indipendent and for example the values of Btot and Tr of R11 will be associated with values of x,y, z corresponding to R11.
I wish my explanation is clear. sorry for my english!
Example of data:


x
y
z
R11 
1
3
1
R12     
1
2
2
R13
1
1
3




Tr
B
R11
0,506697
-1150,00

0,53028
-1070,00

0,601049
-862,00
R12
0,605006
-813

0,630972
-732

0,656938
-666
R13
0,871358
-251,4

0,90447
-233,5

0,937583
-217,2

MaryRam




  • Prev by Date: Re: Finding the real part of a symbolic complex expression
  • Next by Date: Question about ColorFunction
  • Previous by thread: Re: Fourier DFT scaling problem
  • Next by thread: Re: Non-trivial Non- linear fit