R help - Jan 2007 - Controlling size of boxplot when it is added in a plot

Greetings,

	I am trying to add a boxplot to the bottom of a histogram, right  
between the histogram bars and the x axis. Here is the code I am  
using at the moment (the par line is probably not relevant for our  
discussion):

hs <- hist(x, breaks = 20, plot = F)
par(mar = c(3,3,2,1))
hist(x, breaks = 20, main = NULL, ylim = c(-2, max(hs$counts)))
boxplot(x, horizontal = T, axes = T, add = T, at = -1)

The problem is the following. As it is, the boxplot restricts itself  
to the -1 line. I would like it to occupy both the -1 and the -2  
lines ( I guess more generally I would like to control how much  
vertical space the "embedded boxplot" occupies). I tried to set the  
width parameter in the boxplot, but that seemed to have no effect at  
all.

On an OT note, I haven't seen this way of combining a histogram with  
a boxplot (perhaps I haven't looked really hard). I thought it would  
be useful for my students to see them next to each other, to develop  
a feeling for what histograms might correspond to what boxplots. Is  
there perhaps some reason why I should avoid showing those graphs to  
them like that, that I am not aware of? Or just a reason why I  
haven't seen them combined like this much?

TIA

Charilaos Skiadas
Department of Mathematics
Hanover College
P.O.Box 108
Hanover, IN 47243

Charilaos Skiadas wrote:
> Greetings,
> 
> 	I am trying to add a boxplot to the bottom of a histogram, right  
> between the histogram bars and the x axis. Here is the code I am  
> using at the moment (the par line is probably not relevant for our  
> discussion):
> 
> hs <- hist(x, breaks = 20, plot = F)
> par(mar = c(3,3,2,1))
> hist(x, breaks = 20, main = NULL, ylim = c(-2, max(hs$counts)))
> boxplot(x, horizontal = T, axes = T, add = T, at = -1)
> 
> The problem is the following. As it is, the boxplot restricts itself  
> to the -1 line. I would like it to occupy both the -1 and the -2  
> lines ( I guess more generally I would like to control how much  
> vertical space the "embedded boxplot" occupies). I tried to set
the
> width parameter in the boxplot, but that seemed to have no effect at  
> all.
  Try setting the boxwex argument instead:

par(mfrow=c(1,2))

set.seed(54321)
x <- rnorm(100)
hs <- hist(x, breaks = 20, plot = F)
hist(x, breaks = 20, main = NULL, ylim = c(-2, max(hs$counts)))
boxplot(x, horizontal = T, axes = T, add = T, at = -1, boxwex = 1)

hist(x, breaks = 20, main = NULL, ylim = c(-2, max(hs$counts)))
boxplot(x, horizontal = T, axes = T, add = T, at = -1, boxwex = 3)
> On an OT note, I haven't seen this way of combining a histogram with  
> a boxplot (perhaps I haven't looked really hard). I thought it would  
> be useful for my students to see them next to each other, to develop  
> a feeling for what histograms might correspond to what boxplots. Is  
> there perhaps some reason why I should avoid showing those graphs to  
> them like that, that I am not aware of? Or just a reason why I  
> haven't seen them combined like this much?
> 
> TIA
> 
> Charilaos Skiadas
> Department of Mathematics
> Hanover College
> P.O.Box 108
> Hanover, IN 47243
> 
> ______________________________________________
> R-help at stat.math.ethz.ch 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. 
-- 
Chuck Cleland, Ph.D.
NDRI, Inc.
71 West 23rd Street, 8th floor
New York, NY 10010
tel: (212) 845-4495 (Tu, Th)
tel: (732) 512-0171 (M, W, F)
fax: (917) 438-0894

Sorry, meant to send this to the list.

On Jan 14, 2007, at 4:28 PM, Charilaos Skiadas wrote:
> On Jan 14, 2007, at 5:50 AM, Chuck Cleland wrote:
>
>>   Try setting the boxwex argument instead:
>
> Thanks Chuck, that does indeed seem to work pretty well. I'm not  
> quite sure what the best way to determine an appropriate size for  
> the boxplot would be, but the following kind of works, at least for  
> the cases I tried. Though I'm not entirely happy with it. And I'm  
> sure I've made a bunch of errors along the way, that someone more  
> experienced in R could spot easily. Feel free to criticize the  
> code. the boxwex default I guess is probably terribly named. One  
> over it is supposed to be the size of the boxplot over the size of  
> the histogram.
>
> force.odd <- function(x) {
> 	x + 1 - x %% 2;
> }
> boxhist <- function(x, boxwex = 8, ...) {
> 	hs <- hist(x, breaks = 20, plot = F)
> 	space <- force.odd(max(floor(hs$counts / boxwex), 1))
> 	plot(hs, main = NULL, ylim = c(-space, max(hs$counts)), ...)
> 	boxplot(x, horizontal = T, axes = T, add = T, at = -space/2,  
> boxwex = space)
> }
> x <- rweibull(300,1,1); boxhist(x)
>
> Haris
Haris

>>>>> "Charilaos" == Charilaos Skiadas <skiadas at
hanover.edu>
>>>>>     on Sat, 13 Jan 2007 23:12:00 -0500 writes:
    Charilaos> I am trying to add a boxplot to the bottom of a
    Charilaos> histogram, right between the histogram bars and
    Charilaos> the x axis. Here is the code I am using at the
    Charilaos> moment (the par line is probably not relevant for
    Charilaos> our discussion):

    Charilaos> hs <- hist(x, breaks = 20, plot = F)
    Charilaos> par(mar = c(3,3,2,1))
    Charilaos> hist(x, breaks = 20, main = NULL, ylim = c(-2,
max(hs$counts)))
    Charilaos> boxplot(x, horizontal = T, axes = T, add = T, at = -1)

    Charilaos> The problem is the following. 
    Charilaos> ........................

    Charilaos> On an OT note, I haven't seen this way of
    Charilaos> combining a histogram with a boxplot (perhaps I
    Charilaos> haven't looked really hard). I thought it would
    Charilaos> be useful for my students to see them next to
    Charilaos> each other, to develop a feeling for what
    Charilaos> histograms might correspond to what boxplots. Is
    Charilaos> there perhaps some reason why I should avoid
    Charilaos> showing those graphs to them like that, that I am
    Charilaos> not aware of? Or just a reason why I haven't seen
    Charilaos> them combined like this much?

The 'sfsmisc' package (on CRAN) has been containing a function 
called  hist.bxp()   which does these plots,
for a very long time --- actually for almost a longer time than R
exists. It was part of our internal collection of S-plus
"goodies" since 1995.

Martin Maechler, ETH Zurich

One approach is to use the subplot function from the TeachingDemos
package.  I could only get everything to line up properly if I specified
the xlim argument to hist directly and the same value as the ylim
argument to boxplot.

Try this:
> x <- rnorm(100)
> library(TeachingDemos)
> hist(x, breaks = 20, main = NULL, ylim = c(-2,
max(hs$counts)),xlim=c(-3,3))
> subplot( boxplot(x, horizontal=T, ylim=c(-3,3),axes=F ),
par('usr')[1:2], c(par('usr')[3],0))
> box()
I have seen things like this in teaching situations, but not much in the
non-academic world.

-- 
Gregory (Greg) L. Snow Ph.D.
Statistical Data Center
Intermountain Healthcare
greg.snow at intermountainmail.org
(801) 408-8111
 
 
> -----Original Message-----
> From: r-help-bounces at stat.math.ethz.ch 
> [mailto:r-help-bounces at stat.math.ethz.ch] On Behalf Of 
> Charilaos Skiadas
> Sent: Saturday, January 13, 2007 9:12 PM
> To: R-Mailingliste
> Subject: [R] Controlling size of boxplot when it is added in a plot
> 
> Greetings,
> 
> 	I am trying to add a boxplot to the bottom of a 
> histogram, right between the histogram bars and the x axis. 
> Here is the code I am using at the moment (the par line is 
> probably not relevant for our
> discussion):
> 
> hs <- hist(x, breaks = 20, plot = F)
> par(mar = c(3,3,2,1))
> hist(x, breaks = 20, main = NULL, ylim = c(-2, 
> max(hs$counts))) boxplot(x, horizontal = T, axes = T, add = 
> T, at = -1)
> 
> The problem is the following. As it is, the boxplot restricts 
> itself to the -1 line. I would like it to occupy both the -1 
> and the -2 lines ( I guess more generally I would like to 
> control how much vertical space the "embedded boxplot" 
> occupies). I tried to set the width parameter in the boxplot, 
> but that seemed to have no effect at all.
> 
> On an OT note, I haven't seen this way of combining a 
> histogram with a boxplot (perhaps I haven't looked really 
> hard). I thought it would be useful for my students to see 
> them next to each other, to develop a feeling for what 
> histograms might correspond to what boxplots. Is there 
> perhaps some reason why I should avoid showing those graphs 
> to them like that, that I am not aware of? Or just a reason 
> why I haven't seen them combined like this much?
> 
> TIA
> 
> Charilaos Skiadas
> Department of Mathematics
> Hanover College
> P.O.Box 108
> Hanover, IN 47243
> 
> ______________________________________________
> R-help at stat.math.ethz.ch 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.
>