Skip to content

Loess fitting with bisquare

4 messages · Marcin Kozak, Greg Snow, Christian Ritz +1 more

Original
Marcin Kozak
# 
Hi,

It seems there is no straightforward way to carry out in R the loess
fitting with bisquare, as given by William Cleveland in his
"Visualizing Data". Am I right?

Thanks in advance,
Marcin

--
"Build up your weaknesses until they become your strong points" -- Knute Rockne
Greg Snow
# 
The loess function in R uses the tricube weights that are described in my copy of Cleveland, so that may do what you want.  If you really want to do the same general idea but with a different weight function, then it is not that hard to write your own function to do the estimating (but I doubt that you will see much improvement over the standard loess function).
Christian Ritz
# 
Hi,

doing a search in R gives

help.search("loess")
?loess


Look out for the "family" argument in the help page.


Christian
Bert Gunter
# 
If you you use robust fitting (family = "symmetric"), there are (at least)
two sets of weights used (and documented in the Help file): tricube weights
are used to weight the points by their distances from the current point
being estimated; bisquare weights are used in the robust fit to weight by
scaled residual size.

-- Bert Gunter
Genentech

-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On
Behalf Of Greg Snow
Sent: Thursday, January 22, 2009 9:38 AM
To: Marcin Kozak; r-help at r-project.org
Subject: Re: [R] Loess fitting with bisquare

The loess function in R uses the tricube weights that are described in my
copy of Cleveland, so that may do what you want.  If you really want to do
the same general idea but with a different weight function, then it is not
that hard to write your own function to do the estimating (but I doubt that
you will see much improvement over the standard loess function).