
Re: Least square method (regression) for exponential functions
Posted:
Nov 28, 2002 2:39 PM


On 28 Nov 2002, khoh@olzetek.com wrote: >Does anyone know, given a data set (x,y), how to fit a model of the >form >y = a1*exp(b1*x) + a2*exp(b2*x) + a3*exp(b3*x) ? >I want to get the parameters(a1, b1, a2, b2, a3, b3) from fitting the >data set to the above equation. >Please give me the answer(preferably including algorithm or source >code). > >Thanks in advance.
If you are assuming that the error term is distributed as N(0, sig2), I would use PROC NLIN in SAS, probably with the option METHOD = MARQUARDT. It requires writing down the partial derivatives of the function, but that is simple in this case.
Andrzej

