[SciPy-user] Multi-peak fitting
Thu Sep 27 23:25:08 CDT 2007
I'm not familiar with the properties of the Savitzky Golay filter but
one thing to be concerned about in general is introducing a phase bias
into the data by filtering out noise before finding features.
(Some fancy filters supposedly get around
this by basically filtering twice -- once going "forwards" then going
I've posted my small amount of code and some sample
data at my website:
The test example for my abstract feature detection
classes gets right on with fitting a local quadratic to a 'spike' over
an appropriate number of points of the raw noisy data. That's done by
making concrete feature sub-classes in neuro_data.py specific to my
problem. Of course there's always some statistical bias introduced,
and the requires some other assumptions but part of the point is that
I've written more general purpose classes where it's up to you to put
in what assumptions are appropriate for your problem.
It contains a readme with more detailed information, including about how
to make it standalone, and a working example test script
(but only if you are running PyDSTool :). The underlying feature
detection classes (in context.py) are essentially
standalone so you can decouple those from PyDSTool easily at least.
Constructive feedback is welcome, as ever...
More information about the SciPy-user