Re: [S] comparing spline fits with different df?

Samir Mishra (
Thu, 9 Apr 1998 01:43:38 -0400

Hi everyone,

I'm having real problems with this train of thought. Spline methods are
supposed to be non-parametric in nature. Of course it would be an
ultimate solution if there was a perfect "df" that one could define for
every situation. Wouldn't that situation just turn it into a case of
parametric statistics? Would the perfect span be just a reflection of
intuition after scientific and statistical interpolation? I know I sound
very naive with what I've just said, but it's for a reason. I have a
somewhat decent amount of experience with the use of splines, but trying
to define the perfect span without prior knowledge of the data...

Am I shooting out the wrong end here? If so, my apologies...


This message was distributed by To unsubscribe
send e-mail to with the BODY of the
message: unsubscribe s-news