Skip to content

Add function to histogram?

8 messages · Robert Lundqvist, Gabor Grothendieck, Peter Wolf +4 more

Original
Robert Lundqvist
# 
Is there any neat way to add a curve (frequency function or the like) to a
histogram or other plots? I haven't found one yet...

Robert
Gabor Grothendieck
#
On 9/20/05, Robert Lundqvist <Robert.Lundqvist at ltu.se> wrote:
Try this:

set.seed(1)
z <- rnorm(100)
library(MASS)
truehist(z)
lines(density(z))
curve(dnorm, color = "red", add = TRUE)
Peter Wolf
#
Robert Lundqvist wrote:
???

dat<-rnorm(100)
hist(dat,prob=TRUE)
x<-seq(-3.5,3.5,length=100)
y<-dnorm(x)
lines(x,y)

have a look at:  http://cran.at.r-project.org/doc/manuals/R-intro.pdf

Peter Wolf
Romain Francois
# 
Le 21.09.2005 10:00, Peter Wolf a ??crit :
For a frequency polygon, try to work around that piece of code, 
following Peter's notations :

dat <- rnorm(100)
h <- hist(dat,prob=TRUE, border="gray", col="gray90")
diffBreaks <- diff(h$breaks)[1]
xx <- c(h$mids[1]-diffBreaks, h$mids, tail(h$mids,1)+diffBreaks)
yy <- c(0, h$density, 0)
lines(xx, yy, lwd=2)

However, you might prefer (and that's probably wise) the kernel density 
estimator : density() that other people suggested.

Romain
Romain Francois
# 
Le 21.09.2005 10:37, Romain Francois a ??crit :
I just added frequency polygon to the R Graph Gallery, see :
http://addictedtor.free.fr/graphiques/RGraphGallery.php?graph=101

Regards,

Romain
Jim Lemon
#
Robert Lundqvist wrote:
For a general solution, one usually has to scale the density function to 
fit the plot. The simple, but convenient, function rescale() in the 
plotrix package can be used like this:

add.density.curve<-function(y,density=NA) {
  xyrange<-par("usr")
  if(is.na(density)) y.density<-density(y)
  lines(seq(xyrange[1],xyrange[2],length=length(y.density$y)),
   rescale(y.density$y,c(0,xyrange[4])))
}

This only works for histograms, but I am working on a general function 
that will outsmart the habit of some functions of running the plot range 
into negative numbers when these are not possible (i.e. you can't have a 
negative count in a histogram - or a negative density except in the very 
latest physics).

Jim
Vincent Goulet
# 
Le 20 Septembre 2005 09:53, Robert Lundqvist a ??crit??:
The one key thing here is not to forget "prob=TRUE" in the call to hist(), to 
ensure that the y-axis is scaled in proportions, not in frequencies.
Francisco J. Zagmutt
# 
To be more precise, when using hist(prob=T) the y axis shows the densities. 
If you want relative frequencies (proportions) you can use the histogram(x, 
type=) function in the package lattice or write your own function.

Cheers

Francisco