s-news
[Top] [All Lists]

Re: super smoothing

To: AHC van Kampen <a.h.vankampen@amc.uva.nl>
Subject: Re: super smoothing
From: Prof Brian Ripley <ripley@stats.ox.ac.uk>
Date: Tue, 28 Dec 2004 11:25:15 +0000 (GMT)
Cc: s-news@lists.biostat.wustl.edu
In-reply-to: <41D13B92.5090205@amc.uva.nl>
References: <41D13B92.5090205@amc.uva.nl>
Haven't you posted this before recently?

You are asking about the internal code of S-PLUS, and you should really ask Insightful technical support, the only people who have access to that code. You almost certainly didn't get an answer here because we don't know for sure.

(From the Fortran version of supsmu I have, I would deduce that supsmu(span="cv") linearly interpolates between three fixed spans using local weights and does not store the interpolation weights. So the effective span varies from point to point and is not even calculated.)

When you find out, please let this list know.


On Tue, 28 Dec 2004, AHC van Kampen wrote:

I'm using the supsmu( span="cv") function of splus. I use cross-validation (span="cv" option) to determine the optimal span. Is there any possibility to find out which span was finally selected. Or the cross-validation results stored anywhere?

[Two very large signatures removed.]

--
Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

<Prev in Thread] Current Thread [Next in Thread>