[R] Equal confidence interval arrowhead lengths across multiple-paneled lattice plots with free y-scales

[Bob Farmer]

Hi.
I've got a lattice plot with multiple panels and two groups superimposed 
on each panel.  Each panel has an independently scaled y-axis (scales = 
list(relation = "free")).

I've successfully put up 95%CI error bars using panel.arrows (and some 
help from the mailing list).  My question is whether I can "unscale" the 
arrowheads so that they appear to have the same length across all panels.
I recognize from the help-file that the "length" argument in 
panel.arrows is in terms of grid units (which I assume are specific to 
the particular panel).  Can anybody think of a way to set the length so 
that it appears uniform across all panels?

Example code:

   summTable<-data.frame("X" = seq(1,12), "Y" = 
c(1,8,3,6,6,5,7,3,8,1,10,-2),
     "Location" = rep(c("Site1", "Site2"), times = c(6,6)),
     "Species" = rep(c("A", "B"), 6)
     )

#change scale for Site 1
   summTable[1:6,2]<-100*summTable[1:6,2]

#arbitrary confidence intervals
   summTable<-cbind(summTable, "ly" = (summTable$Y*(.98)),
     "uy" = (summTable$Y*1.02)
     )

#panel plotting
   panel.func<-function(x,y,ly,uy,subscripts,...){
     ly<-as.numeric(ly)[subscripts]
     uy<-as.numeric(uy)[subscripts]
     panel.arrows(x,ly,x,uy,
       unit = "native",
       angle = 90,
       length = .25,
       code = 3,
       )
     panel.xyplot(x,y,...)
     }
   xyplot(Y ~ X | Location, groups = Species,
     type = "l",
     data = summTable,
     scales = list(relation = "free"),
     ly = summTable$ly, uy = summTable$uy,
     auto.key = T,
     panel = panel.superpose,
     panel.groups = panel.func
     )

Thank you.

--Bob Farmer
(using R 2.6.0 and lattice 0.16-5)