[S] smoothing non-functional data

David J Cummins (CUMMINS_DAVID_J@lilly.com)
Fri, 17 Jul 1998 08:35:57 -0500

Has anyone tried to smooth data where the underlying curve
is not a function, i.e., for a given value on the x axis there is more
than one corresponding value (in the true curve) on the y axis?
An example of this would be DNA or structures for folded protiens.
One might imagine a curli-cew shaped curve: add noise, generate
data points and take away the true curve -- now try to reconstruct the curve.

One strategy that occurrs to me is to form small sliding windows,
such that within each window the curve is a function; then smooth each
window of data and splice (spline) the results together. The problem with
this is that in practice the true curve is unknown and therefore the correct
window size and placement is very problematic.

I'd be interested in hearing any ideas.

Thank you!

David J. Cummins
Senior Statistician
Eli Lilly & Company

This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news