Skip to content

Plot problems: xlim

8 messages · Ronnie Babigumira, Jacques VESLOT, Ben Bolker +3 more

#
Hi,
Still fresh in R, tried to figure this out, now on my second day running with no luck (and a pile of hair on my desk) so 
I have thrown in the towel and would like to ask for some help.

Here is what I am trying to do. I am trying to plot a distribution, I have 99 points, bound in the range

xlim.min: -0.0173
xlim.max: 0.02103

However, I have a value outside this range (0.2454959) which I would like to add to the plot as a line and to do this I 
use abline(v=0.2454959)

This is what I write

 >xlim = c(-0.02, 0.3)
 >denz <- density(morp)
 >plot.density(denz, xlim = xlim, ylim = c(0,70))
 >hist(morp, freq=F, add= T)
 >abline(v=0.2454959)

Without any options, plot.density spreads out nicely, however, naturally, the line I want to add is not plotted since it 
is well outside the range automatically determined by plot.density hence the need to add xlim however this produces 
something I dont find aesthetically appealing. The plot is squeezed out into a very lean "bell" shape.

So (finally) my question, how can i widen the spread of my plot and yet also be able to add my xline.

Many thanks

Ronnie
#
Please give an example of your data.

set.seed(231) 
morp <- rnorm(20)
range(morp)
[1] -2.311664  1.650254

You can plot 2 histograms, one of them with the extreme value:

par(mfrow=c(2,1))
hist(morp, breaks=10, freq=F)
lines(density(morp))
par(mfrow=c(1,2))
hist(morp, breaks=10, freq=F)
lines(density(morp))
hist(morp, breaks=seq(min(morp), max(morp), length=10), xlim=c(-3, 13), 
freq=F)
lines(density(morp))
abline(v=7.5, lty=3)


Ronnie Babigumira a ??crit :
#
Ronnie Babigumira <rb.glists <at> gmail.com> writes:

   It sounds like you might want to break your axis.
plotrix provides a function to draw the axis break,
but you have to mess around with the data scaling
and axis labels yourself.  See  RSiteSearch("axis 
break"); most of these discussions
are about breaking y axes but the same techniques
apply to the x axis.

  good luck,
    Ben Bolker
#
Many thanks Jacques Veslot and Ben Bolker, Im closer to my goal thanks to Jacques's code and Ben I will follow the leads.

Happy Holidays

Ronnie

Ronnie
Ben Bolker wrote:
#
Using a broken axis is not a good solution to this problem (and there
are very few times that using a broken axis is a good idea)

It sounds like you are trying to compare a reference value to a
distribution.  To do this visually they both need to be on the same
scale, so that you can see the distance between the reference value
and the distribution.

Although it may not be aesthetically pleasing, it is true to the data.

Hadley
#
Hadley
Your point is valid and well taken. However, the break still shows how far off the value is from the distribution (I 
intend to add a note to draw a readers attention to this). Anyhow, like I said, I will try the two ideas shared on the 
list and make a note of what works best for me.

Many thanks

Ronnie
hadley wickham wrote:
#
While I appreciate the availability of axis.break, I agree with Hadley
in this case. I would provide two plots, with and without the special
point. Or just the density plot and some numbers. Broken axes require
interpretation which is often easier to do using numbers, e.g. the
mean or range (exclusive of x.special) and x.special. Sometimes
simple numbers really do provide a better 'picture'.

Peter Ehlers
Ronnie Babigumira wrote:
#
Ronnie Babigumira wrote:
Hi Ronnie,

For only one line, it is probably easiest to stick in an axis break and 
label the line on the x axis. Notice that the position of the line is 
arbitrarily set to be far enough beyond the end of the density curve to 
allow room for the axis break.

testdata<-rnorm(50,sd=0.01)
denz<-density(testdata)
plot(denz,xlim=c(-0.02,0.04),axes=FALSE)
box()
axis(1,at=c(-0.02,0,0.02,0.039),labels=c(-0.02,0,0.02,0.2454959))
abline(v=0.039)
axis.break(1,breakpos=0.037)

This is probably a common enough problem for inclusion in the axis.break 
example.

Jim