Karl,
Thank you for this interesting challenge. It turns out to be reasonably
straightforward.
This code does it for the first example in ?MMC.
## This example is based on the example in ?HH:::MMC
library(HH)
## pairwise comparisons
data(catalystm)
## A. individual plots
## base graphics box plot
boxplot(concent ~ catalyst, data=catalystm,
xlab="catalyst", ylab="concentration", las=1)
## aov
catalystm1.aov <- aov(concent ~ catalyst, data=catalystm)
summary(catalystm1.aov)
## MMC plot
old.omd <- par(omd=c(0,.95,0,1))
catalystm.mmc <- mmc(catalystm1.aov, linfct = mcp(catalyst = "Tukey"))
catalystm.mmc
plot(catalystm.mmc, ry=c(50,58), x.offset=1.8)
## tiebreaker plot
plotMatchMMC(catalystm.mmc$mca, xlabel.print=FALSE)
## B. coordinate scales of all three plots by adjusting par values and
## the frame sizes
def.par <- par(no.readonly = TRUE) # save default, for resetting...
## allocate figure 1 the right 2/3 and top 2/3 of the plotting region
## allocate figure 2 the right 2/3 and bottom 1/3 of the plotting region
## allocate figure 3 the left 1/3 and top 2/3 of the plotting region
frame()
layout(matrix(c(3,1,
0,2), 2, 2, byrow = TRUE),
widths=c(1,2), heights=c(2,1))
## show the regions that have been allocated to each plot
layout.show(3)
omd.old <- par(omd=c(0,.95,0,1))
plot(catalystm.mmc, ry=c(50,59), x.offset=1.8)
usr.mmc <- par()$usr
mar.old <- par(mar=c(5,4,0,2)+.1)
plotMatchMMC(catalystm.mmc$mca, xlabel.print=FALSE)
par(mar.old)
boxplot(concent ~ catalyst, data=catalystm,
xlab="catalyst", ylab="concentration", las=1,
ylim=usr.mmc[3:4], yaxs="i")
## If you have a set of plots that don't fit comfortably into the
## space illustrated here, then you will need to increase the plotting
## device window from the default 7in x 7in because the cex is
## hardwired into the plot.mmc.multicomp function.
On Mon, Jan 14, 2013 at 6:37 PM, Richard M. Heiberger <r...@temple.edu>wrote:
> When the group means are close together, the labels will overlap.
> This overlap is actually informative, indicating that the groups are close.
> For this common case, we provide the tiebreaker function
>
> matchMMC
>
> which is also documented and illustrated on the ?MMC page.
>
> I have never tried to put the data on the same scale. It "should" be easy.
> Let me try, and then post an example.
>
> Rich
>
>
> On Mon, Jan 14, 2013 at 3:53 PM, Karl Ove Hufthammer <k...@huftis.org>wrote:
>
>> må. den 14. 01. 2013 klokka 13.58 (-0500) skreiv Richard M. Heiberger:
>> > Please look at the MMC (Mean-mean Multiple Comparisons) plot in the HH
>> > package. It displays both the means and the differences.
>> >
>> > install.packages("HH") ## if you don't already have it.
>> > library(HH)
>> > ?MMC
>>
>> Thanks for the suggestion. Thats a very interesting and clever way of
>> displaying both means and differences. Its not what I was looking for,
>> though, as it doesnt display the actual data.
>>
>> For the record, heres the syntax for using MMC on the dataset mentioned
>> in my original posting:
>>
>> l.mmc=mmc(l, linfct = mcp(trt = "Tukey"))
>> plot(l.mmc)
>>
>> (It looks best on data where the group means arent very close;
>> otherwise the labels might overlap.)
>>
>> --
>> Karl Ove Hufthammer
>>
>> ______________________________________________
>> R-help@r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide
>> http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
>>
>
>
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
