help understanding box plots

Gracias!

This is really close. I'm just missing the "notches", which meaning
is explained in the help page 
("If the notches of two plots do not overlap then the
          medians are significantly different at the 5 percent level.")
but not their actual definition. 

I presume that they are calculated using a t-value but taking the median
and m.a.d. instead of the mean and var, but this is nowhere specified.
With skewed distributions nothes often
make funny plots.

An example using your function, i.e., bx.example(y=rnorm(300)+runif(300))
should be in the help page! Why not sending it to rhelp?

Agus

Dr. Agustin Lobo
Instituto de Ciencias de la Tierra (CSIC)
Lluis Sole Sabaris s/n
08028 Barcelona SPAIN
tel 34 93409 5410
fax 34 93411 0012
alobo at ija.csic.es


On Fri, 22 Feb 2002, David James wrote:

Hola!

The function below does something close to that.  

"bp.example" <- 
function(y, xlab = "Fraction (p)", ylab = "Quantiles Q(p)", ...)
{
   ##
   ## adapted from Bill Cleveland's Visulizing Data (1993)
   ##
   split.screen(fig = rbind(c(0,2/3, 0, 1), c(2/3, 1, 0, 1)))
   screen(1)
   par(cex=1); par(mar=c(5.1, 4.1, 2, 1))

   q <- quantile(y, c(0.25, 0.5, 0.75))
   y <- sort(y)
   p <- ppoints(y)
   iq <- q[3] - q[1]
   bxp.adj<- q[c(1,3)] + c(-1.5, 1.5)*iq 
   lower.adj<- min(y[y>=bxp.adj[1]])
   upper.adj<- max(y[y<=bxp.adj[2]])

   plot(p, y, xlab=xlab, ylab=ylab, ...)
   u <- par("usr")
   b <- (q[3]-q[1])/.50
   a <- q[1] - .25 * b
   abline(a,b)
   abline(h=q, col = 2, lty=2)
   abline(h=c(lower.adj, upper.adj), col = 3, lty=2)
   cxy <- par("cxy")

   # lower right annotations
   text(x = u[c(2,2)] - cxy[1], 
        y = c(lower.adj,q[1]) + cxy[2]/2,
        c("lower adjecent", "lower quartile (Q1)"), adj=1)
   text(x = u[2]-cxy[1], lower.adj- 0.75*cxy[2], 
        'min(y[y >= Q1 - 1.5 * IQR])', adj=1)

   # upper left annotations
   text(x = u[c(1,1,1)] + cxy[1], 
        y = c(q[-1], upper.adj) + cxy[2]/2,
        c("median (Q2)", "upper quartile (Q3)", "upper adjecent"), adj=0)
   text(x = u[1]+cxy[1], y=upper.adj-0.75*cxy[2],
        'max(y[ y <= Q3 + 1.5 * IQR])', adj=0)
   invisible(list(x = p, y = y, a = a, b = b))

   screen(2)
   ylim <- range(y)
   par(cex=1); par(mar=c(5.1, 1.0, 2, 1))
   boxplot(y, ylim = ylim, axes=F, col=2)
   axis(2, labels=F) 
   box()
   close.screen(all=T)
}

Agustin Lobo wrote:

I've always thought that it would most useful having a
graphic example of boxplot including some text 
pointing to the main features of the boxplot
and that would define and explain these 
features. Perhaps this
could be made a simple function (i.e., boxplot.example())
and this function be included in the help entry. Then
the user would just run boxplot.example() to
see a graphic and commented example. It's more
dificult to understand a text describing
the boxplot function than just seeing a commented
graphic example.


Agus


Dr. Agustin Lobo
Instituto de Ciencias de la Tierra (CSIC)
Lluis Sole Sabaris s/n
08028 Barcelona SPAIN
tel 34 93409 5410
fax 34 93411 0012
alobo at ija.csic.es


On 22 Feb 2002, Peter Dalgaard BSA wrote:

Jay Pfaffman <pfaffman at relaxpc.com> writes:

Another naive stats question.  I'm trying to better understand what
boxplots are telling me.  

I think what I see is the median and the boundaries of the 1st and 3rd
quartiles.  The whiskers represent the range of the data unless there
are points which are outside "range" (default: 1.5) times the distance
from the median to that quartile.  Is that right? 

Not quite. 1.5 times the length of the entire box.

I've read the
documentation for boxplot numerous times, but don't quite understand
it well enough to communicate it to my professor who's helping me with
this project.  (You'll be relieved to know that neither of us fancies
ourself a statistician!)

boxplot.stats.Rd had a typo and got updated recently in the
development and patch versions to read

  \item{coef}{this determines how far the plot ``whiskers'' extend out
    from the box.  If \code{coef} is positive, the whiskers extend to
    the
    most extreme data point which is no more than \code{coef} times
    the length of the box away from the box. A value of zero causes
    the whiskers
    to extend to the data extremes (and no outliers be returned).}

(for some reason this hasn't yet found its way to the online snapshot
manuals in http://stat.ethz.ch/R-alpha/R-devel/doc/html/ and friends.
Martin?)

V&R (p. 122) claims that the hinges are "roughly quartiles," so
perhaps my naive understanding is close enough.

Yes. The exact definition is slightly peculiar, but in compliance with
the original definition by Tukey. So I'm told, anyway.

I've got a relatively small data set (n~=12).  I think it would help
to see the data points plotted on top of the boxplots.  Here's what
I'm doing now:

    par(las=2,ps=14,mar=c(15, 4, 4, 2))
    boxplot(split(ranks,c(1:25)), names=items, notch=T, horizontal=F, add=F)

If I could get the points of each of the 25 variables plotted on top
of the box, that'd be great.

Not sure what you're doing there, but maybe some code like this could
help:

 x1<-rnorm(20)
 x2<-rnorm(20)
 boxplot(list(x1=x1,x2=x2))
 points(cbind(1,x1))
 points(cbind(2,x2))


-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)             FAX: (+45) 35327907
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

-- 
David A. James
Statistics Research, Room 2C-253            Phone:  (908) 582-3082       
Bell Labs, Lucent Technologies              Fax:    (908) 582-3340
Murray Hill, NJ 09794-0636