HELP!!! lsqnonlin fitting problem
in [Matlab]
|
Prev: Simulink: How to generate a sine wave using external source as frequency
Next: Using Matlab GUIs within the HTML framework
From: Chiara LONG on 5 May 2010 18:05 currently i'm workng on the minimization of: sum(x)sum(t)[Observations(x,t)-alfa-Ft-beta*exp(a*x)*Ft^2]^2 there're 2296 observations for x(41)*t(56) where a* is known, the optimization is performed with respect to alfa, beta, Ft, t varying. Ft as a factor value with t varying, should I estimate it as a parameter here(so in total 58 parameters :-S) or there can be a much simpler way to do it? Thank you so so much!!!
From: John D'Errico on 5 May 2010 20:09
"Chiara LONG" <dragenius(a)hotmail.com> wrote in message <hrsq31$ppe$1(a)fred.mathworks.com>... > currently i'm workng on the minimization of: > sum(x)sum(t)[Observations(x,t)-alfa-Ft-beta*exp(a*x)*Ft^2]^2 > > there're 2296 observations for x(41)*t(56) > > where a* is known, the optimization is performed with respect to alfa, beta, Ft, t varying. > > Ft as a factor value with t varying, should I estimate it as a parameter here(so in total 58 parameters :-S) or there can be a much simpler way to do it? > > Thank you so so much!!! You cannot estimate these parameters uniquely. This is why you may be having a problem doing the estimation. It will fail to converge to a consistent value, even if the solver does terminate. John |