Transient fitting/deconvolution
- To: mathgroup at yoda.physics.unc.edu
- Subject: Transient fitting/deconvolution
- From: walker at physics.su.oz.au (Paul Walker)
- Date: Wed, 30 Jun 93 15:13:07 EST
I'm currently experimenting with fitting transient experimental data
(noisy) to a sum of exponentials model. Using the Statistics`NonlinearFit`
package works OK for low noise levels but at even moderate noise levels,
the smaller terms (longer time constants) aren't well-fitted and the
iterative search wanders around and goes out of range, as also discussed
by Simon Chandler recently on this mailgroup. Not exactly
confidence-inspiring when analysing real data!
I've been thinking about using a deconvolution method to filter the noise
and even estimate the parameters, particularly since we also want to
deconvolve the instrument response function from the data. I would
appreciate hearing from anybody who has done something like this or who has
any suggestions for us.
Thanks in anticipation,
Paul Walker
{|} Paul Walker {|}
{|} Dept of Applied Physics {|}
{|} University of Sydney 2006 {|}
{|} AUSTRALIA {|}
{|} {|}
{|} Ph: +61 2 692 3622 {|}
{|} Fax: 660 2903 {|}
{|} e- walker at physics.su.oz.au {|}
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
{ "Learning is a process of invention" }
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~