I want axes that cross

Hello, everybody.

A student asked me a howto question I can't answer.  We want the
length of the drawn axes to fill the full width and height of the
plot, like so:

   |
   |         *
   |           *
   |     *
---|-----------------

However, when we use plot with axes=F and then use the axis commands
to add the axes, they do not cross over each other. We get


   |         *
   |           *
         *
       ----------

The axes are not wide enough to cover the entire range of the data. We
do not want to add a box() command, which is a usual answer for this,
because we don't want to draw on top or the side.

Here's my test case:

x <- rnorm(100)
z <- gl(2,50)
y <- rnorm(100, mean= 1.8*as.numeric(z))

plot(x,y,type="n", axes=F)
points(x,y, pch="$",cex=0.7, col=z)
axis(1, col="green", col.axis="green")
axis(2, col="red", col.axis="red")

Can I cause the 2 axes to "cross" as desired?

The axis help says the axis is clipped at the plot region, but it
seems to get clipped even more narrowly thanthat.

I've been searching in r-help quite a bit and am a little surprised
how difficult this is....

on 02/13/2009 01:25 PM Paul Johnson wrote:

Hello, everybody.

A student asked me a howto question I can't answer.  We want the
length of the drawn axes to fill the full width and height of the
plot, like so:

Paul,

I am guessing that you want:

x <- rnorm(100)
z <- gl(2,50)
y <- rnorm(100, mean= 1.8*as.numeric(z))

plot(x,y,type="n", axes=F)
points(x,y, pch="$",cex=0.7, col=z)
axis(1, col="green", col.axis="green")
axis(2, col="red", col.axis="red")

# Draw the box like an "L" on the bottom and left only
box(bty = "l")


Note that you can specify which sides the 'box' is created upon by using
the 'bty' argument. See ?box for more information.

Also, by default, the axes extend the range of 'x' and 'y'  by 4%. You
can use 'xaxs = i' and 'yaxs = i' in the plot() call to restrict the
axes to the true ranges of 'x' and 'y'.  This would be important, for
example, when you want the lower left hand corner of the plot to be at
exact coordinates such as 0,0.

See ?par for more information.

HTH,

Marc Schwartz

Try defining the argument 'pos' in the axis command-line, like:

x <- rnorm(100)
z <- gl(2,50)
y <- rnorm(100, mean= 1.8*as.numeric(z))

plot(x,y,type="n", axes=F)
points(x,y, pch="$",cex=0.7, col=z)
axis(1, col="green", col.axis="green", pos=0)
axis(2, col="red", col.axis="red", pos=-2)

On Fri, Feb 13, 2009 at 1:51 PM, Marc Schwartz
<marc_schwartz at comcast.net> wrote:

on 02/13/2009 01:25 PM Paul Johnson wrote:

Hello, everybody.

A student asked me a howto question I can't answer.  We want the
length of the drawn axes to fill the full width and height of the
plot, like so:

Paul,

I am guessing that you want:

x <- rnorm(100)
z <- gl(2,50)
y <- rnorm(100, mean= 1.8*as.numeric(z))

plot(x,y,type="n", axes=F)
points(x,y, pch="$",cex=0.7, col=z)
axis(1, col="green", col.axis="green")
axis(2, col="red", col.axis="red")

# Draw the box like an "L" on the bottom and left only
box(bty = "l")


Note that you can specify which sides the 'box' is created upon by using
the 'bty' argument. See ?box for more information.

Thanks, I did not find bty under ?box, but found it under par after
you pointed it out.

That does not get the correct output, however, because the black box
covers over my 2 different colored axes.

Even if I weren't color-conscious, it gives me this:

|
|
|_______

not crossed axes, which I want:

  |
  |
_|______
  |

I'm putting in a seed so we will both see the same things in this example.

set.seed(1233240)
x <- rnorm(100)
z <- gl(2,50)
y <- rnorm(100, mean= 1.8*as.numeric(z))
plot(x,y,type="n", axes=F)
points(x,y, pch="$",cex=0.7, col=z)
axis(1, col="green", col.axis="green")
axis(2, col="red", col.axis="red")

# MS recomends:
# Draw the box like an "L" on the bottom and left only
box(bty = "l")

Also, by default, the axes extend the range of 'x' and 'y'  by 4%. You
can use 'xaxs = i' and 'yaxs = i' in the plot() call to restrict the
axes to the true ranges of 'x' and 'y'.  This would be important, for
example, when you want the lower left hand corner of the plot to be at
exact coordinates such as 0,0.

I would be delighted if the axes really did reach 4% outside the data.
But they don't.  I've seen that same thing you are referring to in the
documentation, but there's something wrong about it, In my example
code, we should see the same thing now I've put in a seed. The axes
are smaller than the data range, not equal to 1.04 times the data
range. I see several observations in the graph that are "off the
charts", they are above the highest value of the y axis, or below the
lowest axis value. Similarly, there are observations smaller than the
low end of the x axis and bigger than the largest x axis value.

The 4% may be the plot region's size, but it is surely not the length
of the axis that is drawn?

See ?par for more information.

HTH,

Marc Schwartz

-- 
Paul E. Johnson
Professor, Political Science
1541 Lilac Lane, Room 504
University of Kansas

On Fri, Feb 13, 2009 at 1:51 PM, Marc Schwartz
<marc_schwartz at comcast.net> wrote:

on 02/13/2009 01:25 PM Paul Johnson wrote:

Hello, everybody.

A student asked me a howto question I can't answer.  We want the
length of the drawn axes to fill the full width and height of the
plot, like so:

Paul,

I am guessing that you want:

x <- rnorm(100)
z <- gl(2,50)
y <- rnorm(100, mean= 1.8*as.numeric(z))

plot(x,y,type="n", axes=F)
points(x,y, pch="$",cex=0.7, col=z)
axis(1, col="green", col.axis="green")
axis(2, col="red", col.axis="red")

# Draw the box like an "L" on the bottom and left only
box(bty = "l")


Note that you can specify which sides the 'box' is created upon by using
the 'bty' argument. See ?box for more information.

Thanks, I did not find bty under ?box, but found it under par after
you pointed it out.

That does not get the correct output, however, because the black box
covers over my 2 different colored axes.

Even if I weren't color-conscious, it gives me this:

|
|
|_______

not crossed axes, which I want:

   |
   |
 _|______
   |

I'm putting in a seed so we will both see the same things in this example.

set.seed(1233240)
x <- rnorm(100)
z <- gl(2,50)
y <- rnorm(100, mean= 1.8*as.numeric(z))
plot(x,y,type="n", axes=F)
points(x,y, pch="$",cex=0.7, col=z)
axis(1, col="green", col.axis="green")
axis(2, col="red", col.axis="red")

# MS recomends:
 # Draw the box like an "L" on the bottom and left only
 box(bty = "l")

Also, by default, the axes extend the range of 'x' and 'y'  by 4%. You
can use 'xaxs = i' and 'yaxs = i' in the plot() call to restrict the
axes to the true ranges of 'x' and 'y'.  This would be important, for
example, when you want the lower left hand corner of the plot to be at
exact coordinates such as 0,0.

I would be delighted if the axes really did reach 4% outside the data.
But they don't.  I've seen that same thing you are referring to in the
documentation, but there's something wrong about it, In my example
code, we should see the same thing now I've put in a seed. The axes
are smaller than the data range, not equal to 1.04 times the data
range. I see several observations in the graph that are "off the
charts", they are above the highest value of the y axis, or below the
lowest axis value. Similarly, there are observations smaller than the
low end of the x axis and bigger than the largest x axis value.

The 4% may be the plot region's size, but it is surely not the length
of the axis that is drawn?

From ?par for 'xaxs':

on 02/13/2009 02:19 PM Paul Johnson wrote:

On Fri, Feb 13, 2009 at 1:51 PM, Marc Schwartz

OK, so given all of the above, something like the following should work:

set.seed(1233240)
x <- rnorm(100)
z <- gl(2,50)
y <- rnorm(100, mean = 1.8 * as.numeric(z))

# Calculate a new range, subtracting a definable value
# from the min of each vector for the new minimum
# Adust the 0.25 as may be needed
X <- c(min(x) - 0.25, max(x))
Y <- c(min(y) - 0.25, max(y))

# Use 'X' and "Y' here, not 'x' and 'y'
# So that the plot region is extended appropriately
plot(X, Y, type = "n", axes = F, xlab = "x", ylab = "y")

points(x, y, pch = "$", cex = 0.7, col = z)

# DO use 'pos'...
axis(1, pos = Y[1], col = "green", col.axis = "green")
axis(2, pos = X[1], col = "red", col.axis = "red")

# get the plot region boundaries
usr <- par("usr")

segments(X[1], usr[3], X[1], usr[4], col = "red")
segments(usr[1], Y[1], usr[2], Y[1], col = "green")


HTH,

Marc

Have you ever seen a drawing of the regions of an R plot with the
terminology that is used for parts?

From what I can remember, several documents on CRAN cover this. The one that

On Fri, Feb 13, 2009 at 3:14 PM, Marc Schwartz
<marc_schwartz at comcast.net> wrote:

on 02/13/2009 02:19 PM Paul Johnson wrote:

On Fri, Feb 13, 2009 at 1:51 PM, Marc Schwartz

OK, so given all of the above, something like the following should work:

set.seed(1233240)
x <- rnorm(100)
z <- gl(2,50)
y <- rnorm(100, mean = 1.8 * as.numeric(z))

# Calculate a new range, subtracting a definable value
# from the min of each vector for the new minimum
# Adust the 0.25 as may be needed
X <- c(min(x) - 0.25, max(x))
Y <- c(min(y) - 0.25, max(y))

# Use 'X' and "Y' here, not 'x' and 'y'
# So that the plot region is extended appropriately
plot(X, Y, type = "n", axes = F, xlab = "x", ylab = "y")

points(x, y, pch = "$", cex = 0.7, col = z)

# DO use 'pos'...
axis(1, pos = Y[1], col = "green", col.axis = "green")
axis(2, pos = X[1], col = "red", col.axis = "red")

# get the plot region boundaries
usr <- par("usr")

segments(X[1], usr[3], X[1], usr[4], col = "red")
segments(usr[1], Y[1], usr[2], Y[1], col = "green")


HTH,

Marc

Thanks, Marc and everybody.

This last suggestion produces the graph I was trying for.

The other approaches that people suggest, using pos or xaxp, produce
other undesired changes in the tick marks or the position of the axes
relative to the data. xaxp offers the promise of a more intuitive
solution, except that, when using it, the tick marks are pushed off in
a bad way.  Your use of segments to draw the extensions of the axes
was the first intuition I had, but I did not know about the trick you
used to retrieve the size of the usr box.

More secrets of par, revealed.

Have you ever seen a drawing of the regions of an R plot with the
terminology that is used for parts?  Until I saw your example code, I
had not understood that the plot axes are placed at the absolute edge
of the user plotting region, for example.

-- 
Paul E. Johnson
Professor, Political Science
1541 Lilac Lane, Room 504
University of Kansas

Hi Paul,

Have you ever seen a drawing of the regions of an R plot with the
terminology that is used for parts?

From what I can remember, several documents on CRAN cover this. The one that

springs to mind is Alzola & Harrell's "An Introduction to S and the Hmisc
and Design Libraries,? which you can download from the Contributed
Documentation link on CRAN.

Chap. 4 of MASS by Venables & Ripley (4th ed.) will also give you what you
want.

Regards, Mark.


Paul Johnson-11 wrote:

On Fri, Feb 13, 2009 at 3:14 PM, Marc Schwartz
<marc_schwartz at comcast.net> wrote:

on 02/13/2009 02:19 PM Paul Johnson wrote:

On Fri, Feb 13, 2009 at 1:51 PM, Marc Schwartz

OK, so given all of the above, something like the following should work:

set.seed(1233240)
x <- rnorm(100)
z <- gl(2,50)
y <- rnorm(100, mean = 1.8 * as.numeric(z))

# Calculate a new range, subtracting a definable value
# from the min of each vector for the new minimum
# Adust the 0.25 as may be needed
X <- c(min(x) - 0.25, max(x))
Y <- c(min(y) - 0.25, max(y))

# Use 'X' and "Y' here, not 'x' and 'y'
# So that the plot region is extended appropriately
plot(X, Y, type = "n", axes = F, xlab = "x", ylab = "y")

points(x, y, pch = "$", cex = 0.7, col = z)

# DO use 'pos'...
axis(1, pos = Y[1], col = "green", col.axis = "green")
axis(2, pos = X[1], col = "red", col.axis = "red")

# get the plot region boundaries
usr <- par("usr")

segments(X[1], usr[3], X[1], usr[4], col = "red")
segments(usr[1], Y[1], usr[2], Y[1], col = "green")


HTH,

Marc

Thanks, Marc and everybody.

This last suggestion produces the graph I was trying for.

The other approaches that people suggest, using pos or xaxp, produce
other undesired changes in the tick marks or the position of the axes
relative to the data. xaxp offers the promise of a more intuitive
solution, except that, when using it, the tick marks are pushed off in
a bad way.  Your use of segments to draw the extensions of the axes
was the first intuition I had, but I did not know about the trick you
used to retrieve the size of the usr box.

More secrets of par, revealed.

Have you ever seen a drawing of the regions of an R plot with the
terminology that is used for parts?  Until I saw your example code, I
had not understood that the plot axes are placed at the absolute edge
of the user plotting region, for example.