How to plot absolute values and differences including confidence intervals

Question

In the followup of the discussion on stackexchange I tried to implement the following plot

from

Cumming, G., & Finch, S. (2005). [Inference by Eye: Confidence Intervals and How to Read Pictures of Data][5]. American Psychologist, 60(2), 170–180. doi:10.1037/0003-066X.60.2.170

I share some people's dislike of double axis, but I think this is a fair use.

Below my partial attempt, the second axis is still missing. I am looking for more elegant alternatives, intelligent variations are welcome.

library(lattice)
library(latticeExtra)
d = data.frame(what=c("A","B","Difference"), 
               mean=c(75,105,30),
               lower=c(50,80,-3),
               upper = c(100,130,63))

# Convert Differences to left scale
d1 = d
d1[d1$what=="Difference",-1] = d1[d1$what=="Difference",-1]+d1[d1=="A","mean"]

segplot(what~lower+upper,centers=mean,data=d1,horizontal=FALSE,draw.bands=FALSE,
        lwd=3,cex=3,ylim=c(0,NA),pch=c(16,16,17),
        panel = function (x,y,z,...){
          centers = list(...)$centers
          panel.segplot(x,y,z,...)
          panel.abline(h=centers[1:2],lty=3)
        } )
## How to add the right scale, close to the last bar?

Answer 1

par(mar=c(3,5,3,5))
plot(NA, xlim=c(.5,3.5), ylim=c(0, max(d$upper[1:2])), bty="l", xaxt="n", xlab="",ylab="Mean")
points(d$mean[1:2], pch=19)
segments(1,d$mean[1],5,d$mean[1],lty=2)
segments(2,d$mean[2],5,d$mean[2],lty=2)
axis(1, 1:3, d$what)
segments(1:2,d$lower[1:2],1:2,d$upper[1:2])
axis(4, seq((d$mean[1]-30),(d$mean[1]+50),by=10), seq(-30,50,by=10), las=1)
points(3,d$mean[1]+d$mean[3],pch=17, cex=1.5)
segments(3,d$lower[3]+d$lower[2],3,d$lower[3]+d$upper[2], lwd=2)
mtext("Difference", side=4, at=d$mean[1], line=3)

Answer 2

As a starting point another base R solution with Hmisc:

library(Hmisc)

with(d1,
     errbar(as.integer(what),mean,upper,lower,xlim=c(0,4),xaxt="n",xlab="",ylim=c(0,150))
     )
points(3,d1[d1$what=="Difference","mean"],pch=15)
axis(1,at=1:3,labels=d1$what)
atics <- seq(floor(d[d$what=="Difference","lower"]/10)*10,ceiling(d[d$what=="Difference","upper"]/10)*10,by=10)
axis(4,at=atics+d1[d1=="A","mean"],labels=atics,pos=3.5)