Skip to content

bug in loess (PR#611)

2 messages · hdrees@statlab.uni-heidelberg.de, Brian Ripley

#
Full_Name: Holger Drees
Version: 1.0.0-1
OS: Windows NT 4.0/Linux Suse 6.3
Submission from: (NULL) (129.206.113.204)


After typing the commands

x <- seq(0,1,by=0.01)
y <- sin(6*x)+rnorm(101)
regtest <- loess(y~x,degree=1,surface="direct",family="symmetric")

regtest$fitted equals NULL. Moreover,

y-regtest$residuals

and 

predict.loess(regtest,x)

give different values, though both are supposed to give the vector of fitted
values.
The latter is the same for both choices of family ("gaussian" and "symmetric").
This holds true for both the WINDOWS and LINUX SUSE 6.3 implementation. (Splus
yields the 
correct results.) This effect occurs only with the option surface="direct".

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
#
On Wed, 19 Jul 2000 hdrees@statlab.uni-heidelberg.de wrote:

            
Since you have clearly looked into this, could you please provide a patch
which gives the correct answers.