C.J.Lee@tnw.utwente.nl ha scritto:
Although I don't do much fitting, I do do a lot with optics and if the background signal you are trying to remove arises from interference fringes then I would fit with a low number of periodic functions.
In fact my procedure would be as follows do an fft on the data and create a power spectrum pick out 1-3 spectral components that describe the fringe pattern Either do a global fit with the frequency components held fixed or fit piece-wise allowing the frequency, phase, and amplitude to vary within some restricted range for that piece.
I tried something like that (first try in fact), but it does not work as smoothly as it can seem. Periodicity is not always perfect too. The polynomial is faster, much easier to implement and works better. By the way, I tried it on > 400 data curves, from different experiments. It seems to work always as expected, despite the (right) concerns. I'll look into splines, but as for now it is well usable this way. m. -- Massimo Sandal University of Bologna Department of Biochemistry "G.Moruzzi" snail mail: Via Irnerio 48, 40126 Bologna, Italy email: massimo.sandal@unibo.it tel: +39-051-2094388 fax: +39-051-2094387