nonlinear regression curve fitting
- To: uiucuxc!yoda.ncsa.uiuc.edu!mathgroup
- Subject: nonlinear regression curve fitting
- From: Walter Vogel - Biochem <uiucuxc!bionette.CGRB.ORST.EDU!vogelw>
- Date: Tue, 7 Nov 89 13:58:09 PST
Does anyone know of a package to do nonlinear regression curve fit to a function, possibly using Marquardt's algorithm?
I am interested in fitting data to an equation of the form y=f(y,x) or
z=f(x,y). Below is the equation I wish to fit.
The equation:
_ _
| |
[L] | F1 F2 |
Y= ----- | ------------------- + ------------------- |
K | 1 + [I]/K1 + [L]/K 1 + [I]/K2 + [L]/K |
|_ _|
There are 4 parameters to be determined: K, K1, K2, F1 (since F1 + F2 =1, F2
is not a parameter. Our raw data is obtained as radioactive counts versus the
independent variable, [I]. The counts are used to calculate both Y and [L] thus
our dependent variable appears on both sides of the equation. Y=(T-n)/r and
[L]=q-T. As written above the equation is a function of two variables [I] and
[L] but it can be recast as function of one T=f(T([I]),[I]).
Sincerely
Walter K. Vogel
vogelw at bionette.cgrb.orst.edu