From: Witold Eryk Wolski
Dear Rgurus,
To my knowledge the best way to visualize the distribution of
a discrete
variable X is
plot(table(X))
The problem which I have is the following. I have to discrete
variables
X and Y which distribution I would like to compare. To overlay the
distribution of Y with lines(table(Y)) gives not satisfying results.
This is the same in case of using density or histogram.
Hence, I am wondering if there is a equivalent of the vioplot
function
(package vioplot) for discrete variables
which starts with a boxplot and than adds a rotated
plot(table()) plot
to each side of the box plot.
Maybee I should ask it first: Does such a plot make any sense? If not
are there better solutions?
cheers
Eryk.
--
Witold Eryk Wolski
__("< School of Mathematics and Statistics _
\__/ University of Newcastle 'v'
|| Newcastle upon Tyne, NE1 7RU, ENGLAND / \
^^ mail: witek96 at users.sourceforge.net m m
Phone : 044 (0)191 222 5376
FAX : 044 (0)191 222 8020
"AndyL" == Liaw, Andy <andy_liaw at merck.com>
on Mon, 21 Mar 2005 08:14:20 -0500 writes:
AndyL> I'd suggest dotcharts, such as:
AndyL> x1 <- sample(letters[1:4], 100, replace=TRUE, prob=c(.2, .3, .4, .1))
AndyL> x2 <- sample(letters[1:4], 100, replace=TRUE, prob=c(.1, .4, .3, .2))
AndyL> f1 <- table(x1) / length(x1)
AndyL> f2 <- table(x2) / length(x2)
AndyL> lev <- factor(c(names(f1), names(f2)))
AndyL> require(lattice)
AndyL> dotplot(lev ~ c(f1, f2), groups=rep(1:2, c(length(f1), length(f2))),
AndyL> panel=panel.superpose)
yes. Maybe slightly even more useful --- and closer to the
plot(table(.)), ...) that Witold mentioned would be the
following slight variation:
dotplot(lev ~ c(f1, f2), groups=rep(1:2, c(length(f1), length(f2))),
panel=panel.superpose, type =c("p","h"))
Note the nice lattice feature (thanks to Deepayan Sarkar!) of
allowing type to be a union of two basic types.
Martin