R help - Oct 2009 - subsetting key on qqmath

Dear R list,

The code below puts qq-plots for two of three groups on the one 
plot.  However the legend includes all three groups, ie the auto.key 
ignores the subset instruction.  Is there an easy way to get around 
this, so that only those groups plotted are included in the legend?

y1<-rnorm(100); y2<-rnorm(100)+1; y3<-rnorm(100)+2; y<-c(y1,y2,y3)
fact=factor(c(rep(1,100),rep(2,100),rep(3,100)))
require(lattice)
qqmath(~y,groups=fact,auto.key=list(corner=c(0,1)),subset=fact!=2)

Many thanks,
John Field


========================John Field Consulting Pty Ltd
10 High Street, Burnside SA 5066 Australia
Phone 08 8332 5294 or 0409 097 586
Fax   08 8332 1229
Email  JohnField at ozemail.com.au 
-------------- next part --------------


Checked by AVG - www.avg.com 

:33:00

On Thu, Oct 8, 2009 at 8:11 PM, John Field <JohnField at ozemail.com.au>
wrote:
> Dear R list,
>
> The code below puts qq-plots for two of three groups on the one plot.
> ?However the legend includes all three groups, ie the auto.key ignores the
> subset instruction. ?Is there an easy way to get around this, so that only
> those groups plotted are included in the legend?
>
> y1<-rnorm(100); y2<-rnorm(100)+1; y3<-rnorm(100)+2;
y<-c(y1,y2,y3)
> fact=factor(c(rep(1,100),rep(2,100),rep(3,100)))
> require(lattice)
> qqmath(~y,groups=fact,auto.key=list(corner=c(0,1)),subset=fact!=2)
The subsetting of groups happens independently in each panel. As
there's no guarantee in general that the subset will have the same
non-empty set of levels in each panel, the only solution (with a
common legend) is to keep all the levels.

To drop unused levels in groups, you need to subset your data
beforehand. For example,

df <- data.frame(y=c(y1,y2,y3),
fact=factor(c(rep(1,100),rep(2,100),rep(3,100))))
qqmath(~y, data=subset(df, fact!=2),
groups=fact[drop=TRUE],auto.key=list(corner=c(0,1)))

-Deepayan