[R] lattice smooth problem?

[Gavin Simpson]

On Sun, 2008-07-06 at 13:52 +0200, Troels Ring wrote:
> Dear friends - I'm on windows, R 2.7.0
> I try again asking if anyone can explain why a single pig of 16 makes so 
> wild swings.
> Warnings are issued, and they are
> 1: pseudoinverse used at 482.1
> 2: neighborhood radius 242.1
> 3: reciprocal condition number  0
> 4: at  360
> 5: radius  14400
> 6: all data on boundary of neighborhood. make span bigger
> 7: There are other near singularities as well. 14400
> 8: pseudoinverse used at 482.1
> 9: neighborhood radius 242.1
> 10: reciprocal condition number  0
> 11: at  360
> 12: radius  14400
> 13: all data on boundary of neighborhood. make span bigger
> 14: There are other near singularities as well. 14400
> 

Did you read the warnings? 6 and 13 both say "make span bigger", and if
one does this:

> xyplot(Na+fit~time|ID,hh,type=c("g","p","smooth"), span = 0.75,
+ auto.key=list(lines=TRUE))

The warnings disappear. The default span is 2/3 which is not big enough
for these data.

HTH

G

> Here is the data and code, ready for copying directly to R
> Best wishes
> Troels
> 
> Na<-c(135,133,131,129,127,126,124,123)
> ID<-c(13,13,13,13,13,13,13,13)
> fit<-c(134.7945,132.9084,131.0848,129.3372,127.5546,125.8162,
> 124.1836,122.6077)
> time<-c(60,120,180,240,300,360,420,480)
> hh<-data.frame(ID=as.factor(ID),Na=Na,fit=fit,time=time)
> library(lattice)
> xyplot(Na+fit~time|ID,hh,type=c("g","p","smooth"),
> auto.key=list(lines=TRUE))
>