Multifactor boxplots

Hi

Dear All

I would like to produce interaction boxplots and this seems to work:

par(mfrow=c(2,2))
A=sample(rnorm(50,50,10))
B=sample(rnorm(50,100,10))
Test=merge(A,B,by=0)#by=0 where 0 is the row.names
TreatA=(gl(2,50,100,labels=c("High","Low")))
TreatB=rep(gl(2,25,50,labels=c("High","Low")),2)
Newdata=data.frame(TreatA,TreatB,Test)

bwplot(x~TreatA:TreatB,data=Newdata)

However, I would prefer the X axis labels to be different, such that 

there

are two label rows (TreatA and TreatB) something like this:

TreatA    High     High    Low       Low
TreatB    High     Low     High      Low

any guidance on achieving this much appreciated.

Combination of axes=FALSE

boxplot(x~TreatA:TreatB,data=Newdata, axes=F)

and following axis and mtext commands seems to do what you want
axis(2)
axis(1, at= 1:4,labels=c("H", "H", "L", "L"))


Regards
Petr

Regards

Tom


   [[alternative HTML version deleted]]

______________________________________________
R-help at 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.

Hi:

Here are a couple of attempts, one with base graphics and one with lattice.

d <- data.frame(TreatA = gl(2, 50, 100, labels = c('High', 'Low')),
                TreatB = rep(gl(2, 25, 50, labels = c('High', 'Low')), 2),
                Test = c(rnorm(50, 50, 10), rnorm(50, 100, 10)))
lbl <- c('High\nHigh', 'High\nLow', 'Low\nHigh', 'Low\nLow')

# Lattice version:
library('lattice')
bwplot(Test ~ interaction(TreatA, TreatB), data = d,
        scales = list(x = list(at = 1:4, label = lbl)))
trellis.focus('toplevel')
panel.text(0.08, 0.065, 'Treat A:\nTreat B:', cex = 0.75)
trellis.unfocus()

boxplot(Test ~ interaction(TreatA, TreatB), data = d, xaxt = 'n')
axis(1, at = 1:4, lab = lbl)
mtext('Treat A:\nTreat B:', side = 1, line = 1, at = 0)

You can adjust the position and size of the text to your taste. The
idea for the lattice plot comes from the following post in the list
archives:
http://r-project.markmail.org/search/?q=panel.text#query:panel.text+page:4+mid:2e3xpb2tsufntwya+state:results

HTH,
Dennis

Dear All

I would like to produce interaction boxplots and this seems to work:

par(mfrow=c(2,2))
A=sample(rnorm(50,50,10))
B=sample(rnorm(50,100,10))
Test=merge(A,B,by=0)#by=0 where 0 is the row.names
TreatA=(gl(2,50,100,labels=c("High","Low")))
TreatB=rep(gl(2,25,50,labels=c("High","Low")),2)
Newdata=data.frame(TreatA,TreatB,Test)

bwplot(x~TreatA:TreatB,data=Newdata)

However, I would prefer the X axis labels to be different, such that there are two label rows (TreatA and TreatB) something like this:

TreatA ? ?High ? ? High ? ?Low ? ? ? Low
TreatB ? ?High ? ? Low ? ? High ? ? ?Low

any guidance on achieving this much appreciated.

Regards

Tom


? ? ? ?[[alternative HTML version deleted]]

______________________________________________
R-help at 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.

Hi

Dear All

I would like to produce interaction boxplots and this seems to work:

par(mfrow=c(2,2))
A=sample(rnorm(50,50,10))
B=sample(rnorm(50,100,10))
Test=merge(A,B,by=0)#by=0 where 0 is the row.names
TreatA=(gl(2,50,100,labels=c("High","Low")))
TreatB=rep(gl(2,25,50,labels=c("High","Low")),2)
Newdata=data.frame(TreatA,TreatB,Test)

bwplot(x~TreatA:TreatB,data=Newdata)

However, I would prefer the X axis labels to be different, such that

there

are two label rows (TreatA and TreatB) something like this:

TreatA    High     High    Low       Low
TreatB    High     Low     High      Low

any guidance on achieving this much appreciated.

Combination of axes=FALSE

boxplot(x~TreatA:TreatB,data=Newdata, axes=F)

and following axis and mtext commands seems to do what you want
axis(2)
axis(1, at= 1:4,labels=c("H", "H", "L", "L"))

I didn't seem to get axes=FALSE to work the way I expected (or even  
find it in boxplot or par documentation... oh, there it is in  
plot.default)  so also tried xaxt="n" and also tried to get the "over/ 
under" arrangement of the labels and "Treat"'s  as specified:

boxplot(x~TreatA:TreatB,data=Newdata, xaxt="n")
axis(1, at= 1:4,labels=c("Treat A:High\nTreat B:High", "High\nLow",  
"Low\nHigh", "Low\nLow"), padj=1 )

---End of answer---


<whine-mode on>
Now I have a question: If axes=FALSE is going to also suppress the  
drawing of the box, shouldn't that be documented in the help page of  
plot.default so that if someone (eventually) finds the right help page  
they can develop the correct expectation? Yes, I can see that  
frame.plot=axes is the default argument, but should we give the useR a  
break and mention that in the Details?

And shouldn't the boxplot help page say that plot.default is  
eventually called and give a link to it so that the available  
graphical parameters are made known to the useR? Well, now that I  
follow the ?bxp link I see that 'axes' is mentioned, but one is  
further referred to ?plot.window which says _nothing_ about 'axes'.  
But then in the code for bxp() I see that plot.default is not called  
and that 'axes' are handled inside that function and nothing is passed  
off to either plot.window or plot.default. <whine-mode off>

David Winsemius, MD
West Hartford, CT