Graphics in R

Things under legendu.net/outdated are outdated technologies that the author does not plan to update any more. Please look for better alternatives.

** Things under legendu.net/outdated are outdated technologies that the author does not plan to update any more. Please look for better alternatives. **

1. par("din") gets the dimension (i.e. width and height) of the graphics device. If no graphics device has been open, it opens a new one.

2. To force a plot to have a fixed aspect ratio, you can use option asp=aspRatio in plotting function. For example, plot(x, y, asp=1) makes a scatter plot with aspect ratio 1. This is helpful when you do not want a plot to get stretched.

3. The function title adds titles and axes labels for an existing plot, so if you forget to specify a title or axes labels when you use produce a plot, you can use title to add them into the plot instead of producing a new plot (sometimes it is very time-consuming to produce a plot, e.g., when you analyze gene data.). legend adds legends for an existing plot. For example, see the code which produces the following image (need to add title).

1. Functions pdf and postscript starts the graphics device driver for producing pdf and ps graphics respectively, which are vectorgraph. Similarly, jpeg, bmp, png and tiff start graphics device driver for producing jpeg, bmp, png and tiff graphics, which are bitmaps.

2. There are several ways to divide a window device to sub window devices (to make plots on them). First, you can use par, e.g. to divide a window device to \(2\times3\) sub window devices, you can use par(mfrow=c(2,3)) or par(mfcol=c(2,3)). The difference between them is the first one display plots by rows while the second one display plots by columns. Second, you can use layout{graphics}, e.g. to divide a window device to two parts vertically and then further divide the bottom part into two horizontal parts, you can use layout(matrix(c(1,2,1,3))). Same values in the matrix stand for the same area, and they also stand for the order they get used among all the sub window devices. Third, you can use split.screen, e.g., to achieve what we did using layout, you can use the following command

split.screen(c(2,1))
split.screen(c(1,2),2)

To make a plot on sub window device generated by split.screen, you must first active it. If you use functions in lattice, you will find that these basically ignore all these 3 functions. lattice has its own way to split window devices. To do this, you have first save plots made by functions in lattice, and then print these plot at appropriate locations using the generic print. For example, the following code make two plots using functions in lattice, and then using print to display them vertically.

px1 = histogram(rnorm(1000))
px2 = bwplot(rnorm(1000))
# arrange the 2 plots vertically
print(px1, position=c(0, 0.6, 1, 1), more=TRUE)
print(px2, position=c(0, 0, 1, 0.4))

The first 2 values and the last 2 (3rd and last) values of the option position stand for x and y coordinates of the bottom-left and top-right corner, respectively, of the panel to display the corresponding plot. Generally speaking, par is most easy one to use, but it can only divide a window device into rectangular sub window devices, and you must make plots in the right order for them to show up at the right places;
layout is also easy to use. You can divide a window device into irregular shapes, but still you have to make plots in the right order to make them to show up at the right place; split.screen is not as convenient as par and layout, but you can divide a window device into irregular sub window devices and make plot on any sub window device (by calling screen first) whenever you want; perhaps it is most inconvenient to use lattice to achieve the same purpose. You must use functions in lattice, and you have to calculate positions for each sub window device and pass it to print. But lattice has some unique plotting functions and it uses more advanced technologies, you can reuse saved plots instead of replotting them every time. This is similar to how Mathematica handles graphics. My suggestion is that you never use lattice for this kind of job unless necessary.

1. Function plot is a generic function (actually many functions in R are generic functions), which be applied to many different types of objects in R.

2. Function boxplot can make plot of different data sets without using function par().

3. Into order to recover to the default graph setting, we can save the information of par first and reuse it whenever we want to go back to the default settings.

4. points add points to a graph; segments add segments to a graph. Though these job can be done by plot, these low level functions are still useful.

5. Mathematica symbols can be used in title and labels of plots using the expression function. For example,

plot(1, xlab=expression(alpha+beta))

1. We can use math expression when plot a graph. To do this, we just need specify the corresponding label as expression(expr), where "expr" is an appropriate expression in R. For example, if we want to use Greek symbol \(\alpha\), we can specify "xlab=expression(alpha)". Whenver we forget how to write mathematica expressions, we can type in command "demo(plotmath)" which will display corresponding expression in R for all kinds of math expressions.

2. sna::gplot.arrow makes neat arrows.

3. plot(..., type = "h") gives us needle plot

4. chull computes convex hull of a given set.

5. dev.copy2pdf copies the graphics content on the current graphics device to a pdf file.

6. You'd better use pdf with dev.off() to make plots, this is a universal way ...
   avoid problems when working on remote servers

ggplot2 Book