R help - Jan 2003 - Plotting coloured histograms...

Hi, I am having some trouble trying to plot a histogram in more than one
colour. What I want to do is, plot two vectors in the same histogram, but
with different colours, for instance:
       > x <- rnorm(1000,20,4);
       > y <- rnorm(1000,10,2);
    Then I'd like to have x and y ploted on the same hist (I can do that
already doing w <- c(x,y) then hist(w)) but the bars representing the x's
should
be in one colour and the bars representing the y should be in another one,
so that I could see the overlaping areas of the two distributions etc.
     Is there any way to do that? I've read through the hist docummentation
(>help(hist)) and also googled for "R colour histogram" but
didn't find
anything helpfull.

Thank you for your attention,

--

Try the H. Bengtsson's function plot.histogram from

http://www.maths.lth.se/matstat/staff/hb/mypackages/R/plot.histogram.R

Remigijus

Saturday, January 25, 2003, 10:21:30 PM, you wrote:

FHFPdR>    Hi, I am having some trouble trying to plot a histogram in more
than one
FHFPdR> colour. What I want to do is, plot two vectors in the same histogram,
but
FHFPdR> with different colours, for instance:
FHFPdR>        > x <- rnorm(1000,20,4);
FHFPdR>        > y <- rnorm(1000,10,2);
FHFPdR>     Then I'd like to have x and y ploted on the same hist (I can
do that
FHFPdR> already doing w <- c(x,y) then hist(w)) but the bars representing
the x's should
FHFPdR> be in one colour and the bars representing the y should be in another
one,
FHFPdR> so that I could see the overlaping areas of the two distributions
etc.
FHFPdR>      Is there any way to do that? I've read through the hist
docummentation
(>>help(hist)) and also googled for "R colour histogram" but
didn't find
FHFPdR> anything helpfull.

FHFPdR> Thank you for your attention,

FHFPdR> --

FHFPdR> ______________________________________________
FHFPdR> R-help at stat.math.ethz.ch mailing list
FHFPdR> http://www.stat.math.ethz.ch/mailman/listinfo/r-help

Dear R helpers,

What are the supported distributions in R and where to look for them?. Also,
is it possible to generate cyclic arrivals in R. For example, arrival
patterns that
correlates with time of the day.

I really appreciate your help,

On Sun, 26 Jan 2003, Mohamed A. Kerasha wrote:
> What are the supported distributions in R and where to look for them?. 
help.search("distribution")

and look also at packages SuppDists, evd, gld and sn amongst others.
> Also,
> is it possible to generate cyclic arrivals in R. For example, arrival
> patterns that
> correlates with time of the day.
Yes, it is possible, provided you have a complete description.  R is after 
all a fully-fledged programming language.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

Hi All,

I plotted several distribution curves in one graph but
I do not know how to add legends say which is what
distribution.  Could someone help me?  Thanks,

Hi,

----- Original Message ----- 
From: "Qin Xin" <qinxin2001 at yahoo.com>
To: <r-help at stat.math.ethz.ch>
Sent: Monday, January 27, 2003 9:42 AM
Subject: [R] how could I add legends?

> Hi All,
> 
> I plotted several distribution curves in one graph but
> I do not know how to add legends say which is what
> distribution.  Could someone help me?  Thanks,

Have a look at
   ?legend

Kevin
------------------------------------------------
Ko-Kang Kevin Wang
Master of Science (MSc) Student
Department of Statistics
University of Auckland
New Zealand
www.stat.auckland.ac.nz/~kwan022

> Hi, I am having some trouble trying to plot a histogram in more than one
> colour. What I want to do is, plot two vectors in the same histogram, but
> with different colours, for instance:
>        > x <- rnorm(1000,20,4);
>        > y <- rnorm(1000,10,2);
>     Then I'd like to have x and y ploted on the same hist (I can do
that
> already doing w <- c(x,y) then hist(w)) but the bars representing the
x's should
> be in one colour and the bars representing the y should be in another one,
> so that I could see the overlaping areas of the two distributions etc.
You haven't made it clear if you want the histograms on top of each other
(which is what you get from w<-c(x,y) ) or if you want them side-by-side
(so you can see overlapping areas).

I plot histograms on top of each other using my own trellis function
panel.surpose, like this:

x <- rnorm(1000,20,4)
y <- rnorm(1000,10,2)
w <- c(x,y)
f <- ifelse(1:length(z)<=length(x),0,1)
library(lattice)
histogram(~w, groups=f, panel=panel.surpose)

The "groups" argument should be a factor, or some other variable
taking
values in a small set. The function panel.surpose takes each panel in
turn, splits the dataset for that panel into groups according to
"groups",
and plots a histogram for each group, one on top of another. The
colours can be specified like "bar.col=c(rgb(1,0,0),rgb(0,1,0))", the
first colour being used for the lowest level of "groups". You could
specify a different "panel.groups" function, but I can't see how
that
would be useful.

In writing this function, I had trouble with the standard routines
hist and panel.histogram, which deal with the counting. What should
their behaviour be when the data has NA values? It seems to me that
count, percentage and density do not handle NA values in a consistent way.
Anyway, I wrote a replacement for panel.histogram to use with my
panel.surpose.

Damon Wischik.


panel.histogram.partial <- 
function (x, breaks, equal.widths = TRUE, type = "density", col
bar.fill$col, ...)
{
    x <- as.numeric(x)
    grid.lines(x = c(0.05, 0.95), y = unit(c(0, 0), "native"), 
        default.units = "npc")
    if (length(x) > 0) {
        bar.fill <- trellis.par.get("bar.fill")
        if (is.null(breaks)) {
            nint <- round(log2(length(x)) + 1)
            breaks <- if (equal.widths) 
                do.breaks(range(x), nint)
            else quantile(x, 0:nint/nint)
        }
        h <- hist(x, breaks = breaks, plot = FALSE, ...)
        y <- if (type == "count") 
            h$counts
        else if (type == "percent") 
            100 * h$counts/length(x)
        else h$intensities * length(which(!is.na(x)))/length(x)
        nb <- length(breaks)
        if (nb != (length(y) + 1)) 
            warning("something is probably wrong")
        if (nb > 1) {
            for (i in 1:(nb - 1)) if (y[i] > 0) {
                grid.rect(gp = gpar(fill = col), x = breaks[i], 
                  y = 0, height = y[i], width = breaks[i + 1] - 
                    breaks[i], just = c("left", "bottom"),
default.units="native")
            }
        }
    }
}

rgbmix <- function(p,col1,col2) { # p a vector. pi=0: col1. pi=1: col2
  cm <- (1-p) %o% col2rgb(col1) + p %o% col2rgb(col2)
 
rgb(cm[,"red",],cm[,"green",],cm[,"blue",],maxColorValue=255)
  }

panel.surpose <-
function (x, y = NULL, subscripts, groups, panel.groups
"panel.histogram.partial",
  bar.col=superpose.line$col, ...)
  {
  x <- as.numeric(x)
  if (!is.null(y)) y <- as.numeric(y)
  superpose.line <- trellis.par.get("superpose.line")
  if (length(subscripts)>0 && length(groups)<max(subscripts))
    groups <- rep(groups, length=max(subscripts))
  vals <- sort(unique(groups))
  nvals <- length(vals)
  bar.col <- if (length(bar.col)>=nvals) bar.col else {
    if (length(bar.col)>2 || length(bar.col)==1) 
      rep(bar.col,length=nvals)
    else # linearly interpolate the colours
      rgbmix((1:nvals-1)/(nvals-1),bar.col[[1]],bar.col[[2]])
    }
  panel.groups <- if(is.function(panel.groups)) 
    panel.groups
  else if (is.character(panel.groups))
    get(panel.groups)
  else
    eval(panel.groups)
  nx <- length(subscripts)
  for (i in nvals:1) {
    id <- (groups[subscripts] %in% vals[1:i])
    if (any(id)) {
      fakex <- rep(NA,times=nx-length(which(id)))
      xandfake <- c(fakex,x[id])
      args <- if (!is.null(y))
        list(x=xandfake,y,col=bar.col[i],...)
      else
        list(x=xandfake, col=bar.col[i], lty=0, ...)
      do.call("panel.groups",args)
      }
    }
  }

On 25 Jan 2003 at 22:21, Fernando Henrique Ferraz Pereira da Rosa 
wrote:

Hola!

Maybe this may be of help:
> x <- rnorm(100,2,2)
> y <- rnorm(200, 4,3)
> hist(x)
> hist(y, add=TRUE)
> # which gives a confusing result. Better is:
> hist(x, freq=FALSE)
> hist(y, add=TRUE, freq=FALSE)
> # But it is difficult to separate visually the two histograms in
> # the same plot. Trying with colors:
> hist(x, freq=FALSE, col="blue")
> hist(y, add=TRUE, freq=FALSE, col="red")
> # is somewhat better, but it covers part of the blue histogram!
> # Anybody knows of better solutions?
Kjetil Halvorsen

>    Hi, I am having some trouble trying to plot a histogram in more than one
> colour. What I want to do is, plot two vectors in the same histogram, but
> with different colours, for instance:
>        > x <- rnorm(1000,20,4);
>        > y <- rnorm(1000,10,2);
>     Then I'd like to have x and y ploted on the same hist (I can do
that
> already doing w <- c(x,y) then hist(w)) but the bars representing the
x's should
> be in one colour and the bars representing the y should be in another one,
> so that I could see the overlaping areas of the two distributions etc.
>      Is there any way to do that? I've read through the hist
docummentation
> (>help(hist)) and also googled for "R colour histogram" but
didn't find
> anything helpfull.
> 
> Thank you for your attention,
> 
> --
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> http://www.stat.math.ethz.ch/mailman/listinfo/r-help