You will definitely be able to run the code and the examples shown in this book directly from the standard R Graphical User Interface (GUI), especially if you are frequently working with R in more complex projects or simply if you like to keep an eye on the different components of your code, such as scripts, plots, and help pages, you may well think about the possibility of using an IDE. The number of specific IDEs that get integrated with R is still limited, but some of them are quite efficient, well-designed and open source.

RStudio

RStudio (http://www.rstudio.com/) is a very nice and advanced programming environment developed specifically for R, and this would be my recommended choice of IDE as the R programming environment in most cases. It is available for all the major platforms (Windows, Linux, and Mac OS X), and it can be run on a local machine, such as your computer, or even over the Web, using RStudio Server. With RStudio Server, you can connect a browser-based interface (the RStudio IDE) to a version of R running on a remote Linux server.

RStudio allows you to integrate several useful functionalities, in particular if you use R for a more complex project. The way the software interface is organized allows you to keep an eye on the different activities you very often deal with in R, such as working on different scripts, overviewing the installed packages, as well as having easy access to the help pages and the plots generated. This last feature is particularly interesting for ggplot2 since in RStudio, you will be able to easily access the history of the plots created instead of visualizing only the last created plot, as is the case in the default R GUI. One other very useful feature of RStudio is code completion. You can, in fact, start typing a comment, and upon pressing the Tab key, the interface will provide you with functions matching what you have written . This feature will turn out to be very useful in ggplot2, so you will not necessarily need to remember all the functions and you will also have guidance for the arguments of the functions as well.

In Figure 1.1, you can see a screenshot from the current version of RStudio (v 0.98.1091):

Figure 1.1: This is a screenshot of RStudio on Windows 8

The environment is composed of four different areas:

• Scripting area: In this area you can open, create, and write the scripts.
• Console area: This area is the actual R console in which the commands are executed. It is possible to type commands directly here in the console or write them in a script and then run them on the console (I would recommend the last option).
• Workspace/History area: In this area, you can find a practical summary of all the objects created in the workspace in which you are working and the history of the typed commands.
• Visualization area: Here, you can easily load packages, open R help files, and, even more importantly, visualize plots.

The RStudio website provides a lot of material on how to use the program, such as manuals, tutorials, and videos, so if you are interested, refer to the website for more details.

R provides a complete series of options to realize graphics, which makes it quite advanced with regard to data visualization. Along the next few sections of this chapter, we will go through the most important R packages for data visualization by quickly discussing some high-level differences and analogies. If you already have some experience with other R packages for data visualization, in particular graphics or lattice, the following sections will provide you with some references and examples of how the code used in such packages appears in comparison with that used in ggplot2. Moreover, you will also have an idea of the typical layout of the plots created with a certain package, so you will be able to identify the tool used to realize the plots you will come across.

The core of graphics visualization in R is within the grDevices package, which provides the basic structure of data plotting, such as the colors and fonts used in the plots. Such a graphic engine was then used as the starting point in the development of more advanced and sophisticated packages for data visualization, the most commonly used being graphics and grid.

The graphics package is often referred to as the base or traditional graphics environment since, historically, it was the first package for data visualization available in R, and it provides functions that allow the generation of complete plots.

The grid package, on the other hand, provides an alternative set of graphics tools. This package does not directly provide functions that generate complete plots, so it is not frequently used directly to generate graphics, but it is used in the development of advanced data visualization packages. Among the grid-based packages, the most widely used are lattice and ggplot2, although they are built by implementing different visualization approaches—Trellis plots in the case of lattice and the grammar of graphics in the case of ggplot2. We will describe these principles in more detail in the coming sections. A diagram representing the connections between the tools just mentioned is shown in Figure 1.2. Just keep in mind that this is not a complete overview of the packages available but simply a small snapshot of the packages we will discuss. Many other packages are built on top of the tools just mentioned, but in the following sections, we will focus on the most relevant packages used in data visualization, namely graphics, lattice, and, of course, ggplot2. If you would like to get a more complete overview of the graphics tools available in R, you can have a look at the web page of the R project summarizing such tools, http://cran.r-project.org/web/views/Graphics.html.

Figure 1.2: This is an overview of the most widely used R packages for graphics

In order to see some examples of plots in graphics, lattice and ggplot2, we will go through a few examples of different plots over the following pages. The objective of providing these examples is not to do an exhaustive comparison of the three packages but simply to provide you with a simple comparison of how the different codes as well as the default plot layouts appear for these different plotting tools. For these examples, we will use the Orange dataset available in R; to load it in the workspace, simply write the following code:

>data(Orange)

This dataset contains records of the growth of orange trees. You can have a look at the data by recalling its first lines with the following code:

>head(Orange)

You will see that the dataset contains three columns. The first one, Tree, is an ID number indicating the tree on which the measurement was taken, while age and circumference refer to the age in days and the size of the tree in millimeters, respectively. If you want to have more information about this data, you can have a look at the help page of the dataset by typing the following code:

?Orange

Here, you will find the reference of the data as well as a more detailed description of the variables included.

The existence of these two different graphics environments brings these questions to most users' minds—which package to use and under which circumstances? For simple and basic plots, where the data simply needs to be represented in a standard plot type (such as a scatter plot, histogram, or boxplot) without any additional manipulation, then all the plotting environments are fairly equivalent. In fact, it would probably be possible to produce the same type of plot with graphics as well as with lattice or ggplot2. Nevertheless, in general, the default graphic output of ggplot2 or lattice will be most likely superior compared to graphics since both these packages are designed considering the principles of human perception deeply and to make the evaluation of data contained in plots easier.

When more complex data should be analyzed, then the grid-based packages, lattice and ggplot2, present a more sophisticated support in the analysis of multivariate data. On the other hand, these tools require greater effort to become proficient because of their flexibility and advanced functionalities. In both cases, lattice and ggplot2, the package provides a full set of tools for data visualization, so you will not need to use grid directly in most cases, but you will be able to do all your work directly with one of those packages.

The graphics package was originally developed based on the experience of the graphics environment in R. The approach implemented in this package is based on the principle of the pen-on-paper model, where the plot is drawn in the first function call and once content is added, it cannot be deleted or modified.

In general, the functions available in this package can be divided into high-level and low-level functions. High-level functions are functions capable of drawing the actual plot, while low-level functions are functions used to add content to a graph that was already created with a high-level function.

Let's assume that we would like to have a look at how age is related to the circumference of the trees in our dataset Orange; we could simply plot the data on a scatter plot using the high-level function plot() as shown in the following code:

plot(age~circumference, data=Orange)

This code creates the graph in Figure 1.3. As you would have noticed, we obtained the graph directly with a call to a function that contains the variables to plot in the form of y~x, and the dataset to locate them. As an alternative, instead of using a formula expression, you can use a direct reference to x and y, using code in the form of plot(x,y). In this case, you will have to use a direct reference to the data instead of using the data argument of the function. Type in the following code:

plot(Orange$circumference, Orange$age)

The preceding code results in the following output:

Figure 1.3: Simple scatterplot of the dataset Orange using graphics

For the time being, we are not interested in the plot's details, such as the title or the axis, but we will simply focus on how to add elements to the plot we just created. For instance, if we want to include a regression line as well as a smooth line to have an idea of the relation between the data, we should use a low-level function to add the just-created additional lines to the plot; this is done with the lines() function:

plot(age~circumference, data=Orange)   ###Create basic plot
abline(lm(Orange$age~Orange$circumference), col="blue")
lines(loess.smooth(Orange$circumference,Orange$age), col="red")

The graph generated as the output of this code is shown in Figure 1.4:

Figure 1.4: This is a scatterplot of the Orange data with a regression line (in blue) and a smooth line (in red) realized with graphics

As illustrated, with this package, we have built a graph by first calling one function, which draws the main plot frame, and then additional elements were included using other functions. With graphics, only additional elements can be included in the graph without changing the overall plot frame defined by the plot() function. This ability to add several graphical elements together to create a complex plot is one of the fundamental elements of R, and you will notice how all the different graphical packages rely on this principle. If you are interested in getting other code examples of plots in graphics, there is also some demo code available in R for this package, and it can be visualized with demo(graphics).

In the coming sections, you will find a quick reference to how you can generate a similar plot using graphics and ggplot2. As will be described in more detail later on, in ggplot2, there are two main functions to realize plots, ggplot() and qplot(). The function qplot() is a wrapper function that is designed to easily create basic plots with ggplot2, and it has a similar code to the plot() function of graphics. Due to its simplicity, this function is the easiest way to start working with ggplot2, so we will use this function in the examples in the following sections. The code in these sections also uses our example dataset Orange; in this way, you can run the code directly on your console and see the resulting output.

Scatterplots with individual data points

To generate the plot generated using graphics, use the following code:

plot(age~circumference, data=Orange)

The preceding code results in the following output:

To generate the plot using ggplot2, use the following code:

qplot(circumference,age, data=Orange)

The preceding code results in the following output:

Scatterplots with the line of one tree

To generate the plot using graphics, use the following code:

plot(age~circumference, data=Orange[Orange$Tree==1,], type="l")

The preceding code results in the following output:

To generate the plot using ggplot2, use the following code:

qplot(circumference,age, data=Orange[Orange$Tree==1,], geom="line")

The preceding code results in the following output:

Scatterplots with the line and points of one tree

To generate the plot using graphics, use the following code:

plot(age~circumference, data=Orange[Orange$Tree==1,], type="b")

The preceding code results in the following output:

To generate the plot using ggplot2, use the following code:

qplot(circumference,age, data=Orange[Orange$Tree==1,], geom=c("line","point"))

The preceding code results in the following output:

Boxplots of the orange dataset

To generate the plot using graphics, use the following code:

boxplot(circumference~Tree, data=Orange)

The preceding code results in the following output:

To generate the plot using ggplot2, use the following code:

qplot(Tree,circumference, data=Orange, geom="boxplot")

The preceding code results in the following output:

Boxplots with individual observations

To generate the plot using graphics, use the following code:

boxplot(circumference~Tree, data=Orange)
points(circumference~Tree, data=Orange)

The preceding code results in the following output:

To generate the plot using ggplot2, use the following code:

qplot(Tree,circumference, data=Orange, geom=c("boxplot","point"))

The preceding code results in the following output:

Histograms of the orange dataset

To generate the plot using graphics, use the following code:

hist(Orange$circumference)

The preceding code results in the following output:

To generate the plot using ggplot2, use the following code:

qplot(circumference, data=Orange, geom="histogram")

The preceding code results in the following output:

Histograms with the reference line at the median value in red

To generate the plot using graphics, use the following code:

hist(Orange$circumference)
abline(v=median(Orange$circumference), col="red")

The preceding code results in the following output:

To generate the plot using ggplot2, use the following code:

qplot(circumference, data=Orange, geom="histogram")+geom_vline(xintercept = median(Orange$circumference), colour="red")

The preceding code results in the following output:

Along with with graphics, the base R installation also includes the lattice package. This package implements a family of techniques known as Trellis graphics, proposed by William Cleveland to visualize complex datasets with multiple variables. The objective of those design principles was to ensure the accurate and faithful communication of data information. These principles are embedded into the package and are already evident in the default plot design settings. One interesting feature of Trellis plots is the option of multipanel conditioning, which creates multiple plots by splitting the data on the basis of one variable. A similar option is also available in ggplot2, but in that case, it is called faceting.

In lattice, we also have functions that are able to generate a plot with one single call, but once the plot is drawn, it is already final. Consequently, plot details as well as additional elements that need to be included in the graph, need to be specified already within the call to the main function. This is done by including all the specifications in the panel function argument. These specifications can be included directly in the main body of the function or specified in an independent function, which is then called; this last option usually generates more readable code, so this will be the approach used in the following examples. For instance, if we want to draw the same plot we just generated in the previous section with graphics, containing the age and circumference of trees and also the regression and smooth lines, we need to specify such elements within the function call. You may see an example of the code here; remember that lattice needs to be loaded in the workspace:

require(lattice)              ##Load lattice if needed
myPanel <- function(x,y){
panel.xyplot(x,y)            # Add the observations 
panel.lmline(x,y,col="blue")   # Add the regression
panel.loess(x,y,col="red")      # Add the smooth line
}
xyplot(age~circumference, data=Orange, panel=myPanel)

This code produces the plot in Figure 1.5:

Figure 1.5: This is a scatter plot of the Orange data with the regression line (in blue) and the smooth line (in red) realized with lattice

As you would have noticed, taking aside the code differences, the plot generated does not look very different from the one obtained with graphics. This is because we are not using any special visualization feature of lattice. As mentioned earlier, with this package, we have the option of multipanel conditioning, so let's take a look at this. Let's assume that we want to have the same plot but for the different trees in the dataset. Of course, in this case, you would not need the regression or the smooth line, since there will only be one tree in each plot window, but it could be nice to have the different observations connected. This is shown in the following code:

myPanel <- function(x,y){
panel.xyplot(x,y, type="b") #the observations
}
xyplot(age~circumference | Tree, data=Orange, panel=myPanel)

This code generates the graph shown in Figure 1.6:

Figure 1.6: This is a scatterplot of the Orange data realized with lattice, with one subpanel representing the individual data of each tree. The number of trees in each panel is reported in the upper part of the plot area

As illustrated, using the vertical bar |, we are able to obtain the plot conditional to the value of the variable Tree. In the upper part of the panels, you would notice the reference to the value of the conditional variable, which, in this case, is the column Tree. As mentioned before, ggplot2 offers this option too; we will see one example of that in the next section.

In the next section, You would find a quick reference to how to convert a typical plot type from lattice to ggplot2. In this case, the examples are adapted to the typical plotting style of the lattice plots.

Scatterplots with individual observations

To plot the graph using lattice, use the following code:

xyplot(age~circumference, data=Orange)

The preceding code results in the following output:

To plot the graph using ggplot2, use the following code:

qplot(circumference,age, data=Orange)

The preceding code results in the following output:

Scatterplots of the orange dataset with faceting

To plot the graph using lattice, use the following code:

xyplot(age~circumference|Tree, data=Orange)

The preceding code results in the following output:

To plot the graph using ggplot2, use the following code:

qplot(circumference,age, data=Orange, facets=~Tree)

The preceding code results in the following output:

Faceting scatterplots with line and points

To plot the graph using lattice, use the following code:

xyplot(age~circumference|Tree, data=Orange, type="b")

The preceding code results in the following output:

To plot the graph using ggplot2, use the following code:

qplot(circumference,age, data=Orange, geom=c("line","point"), facets=~Tree)

The preceding code results in the following output:

Scatterplots with grouping data

To plot the graph using lattice, use the following code:

xyplot(age~circumference, data=Orange, groups=Tree, type="b")

The preceding code results in the following output:

To plot the graph using ggplot2, use the following code:

qplot(circumference,age, data=Orange,color=Tree, geom=c("line","point"))

The preceding code results in the following output:

Boxplots of the orange dataset

To plot the graph using lattice, use the following code:

bwplot(circumference~Tree, data=Orange)

The preceding code results in the following output:

To plot the graph using ggplot2, use the following code:

qplot(Tree,circumference, data=Orange, geom="boxplot")

The preceding code results in the following output:

Histograms of the orange dataset

To plot the graph using lattice, use the following code:

histogram(Orange$circumference, type = "count")

To plot the graph using ggplot2, use the following code:

qplot(circumference, data=Orange, geom="histogram")

The preceding code results in the following output:

Histograms with the reference line at the median value in red

To plot the graph using lattice, use the following code:

histogram(~circumference, data=Orange, type = "count", panel=function(x,...){panel.histogram(x, ...);panel.abline(v=median(x), col="red")})

The preceding code results in the following output:

To plot the graph using ggplot2, use the following code:

qplot(circumference, data=Orange, geom="histogram")+geom_vline(xintercept = median(Orange$circumference), colour="red")

The preceding code results in the following output:

The ggplot2 package was developed by Hadley Wickham by implementing a completely different approach to statistical plots. As is the case with lattice, this package is also based on grid, providing a series of high-level functions that allow the creation of complete plots. The ggplot2 package provides an interpretation and extension of the principles of the book The Grammar of Graphics by Leland Wilkinson. Briefly, The Grammar of Graphics assumes that a statistical graphic is a mapping of data to the aesthetic attributes and geometric objects used to represent data, such as points, lines, bars, and so on. Besides the aesthetic attributes, the plot can also contain statistical transformation or grouping of data. As in lattice, in ggplot2, we have the possibility of splitting data by a certain variable, obtaining a representation of each subset of data in an independent subplot; such representation in ggplot2 is called faceting.

In a more formal way, the main components of the grammar of graphics are the data and its mapping, aesthetics, geometric objects, statistical transformations, scales, coordinates, and faceting. We will cover each one of these elements in more detail in Chapter 3, The Layers and Grammar of Graphics, but for now, consider these general principles:

In ggplot2, there are two main high-level functions capable of directly creating a plot, qplot(), and ggplot(); qplot() stands for quick plot, and it is a simple function that serves a purpose similar to that served by the plot() function in graphics. The ggplot()function, on the other hand, is a much more advanced function that allows the user to have more control of the plot layout and details. In our journey into the world of ggplot2, we will see some examples of qplot(), in particular when we go through the different kinds of graphs, but we will dig a lot deeper into ggplot() since this last function is more suited to advanced examples.

If you have a look at the different forums based on R programming, there is quite a bit of discussion as to which of these two functions would be more convenient to use. My general recommendation would be that it depends on the type of graph you are drawing more frequently. For simple and standard plots, where only the data should be represented and only the minor modification of standard layouts are required, the qplot() function will do the job. On the other hand, if you need to apply particular transformations to the data or if you would just like to keep the freedom of controlling and defining the different details of the plot layout, I would recommend that you focus on ggplot(). As you will see, the code between these functions is not completely different since they are both based on the same underlying philosophy, but the way in which the options are set is quite different, so if you want to adapt a plot from one function to the other, you will essentially need to rewrite your code. If you just want to focus on learning only one of them, I would definitely recommend that you learn ggplot().

In the following code, you will see an example of a plot realized with ggplot2, where you can identify some of the components of the grammar of graphics. The example is realized with the ggplot() function, which allows a more direct comparison with the grammar of graphics, but coming just after the following code, you could also find the corresponding qplot() code useful. Both codes generate the graph depicted in Figure 1.7:

require(ggplot2)                             ## Load ggplot2
data(Orange)                                 ## Load the data

ggplot(data=Orange,                          ## Data used
  aes(x=circumference,y=age, color=Tree))+   ## Aesthetic
geom_point()+                                ## Geometry 
stat_smooth(method="lm",se=FALSE)            ## Statistics

### Corresponding code with qplot()
qplot(circumference,age,data=Orange,         ## Data used
  color=Tree,                                ## Aesthetic mapping 
  geom=c("point","smooth"),method="lm",se=FALSE)

This simple example can give you an idea of the role of each portion of code in a ggplot2 graph; you have seen how the main function body creates the connection between the data and the aesthetics we are interested to represent and how, on top of this, you add the components of the plot, as in this case, we added the geometry element of points and the statistical element of regression. You can also notice how the components that need to be added to the main function call are included using the + sign. One more thing worth mentioning at this point is that if you run just the main body function in the ggplot() function, you will get an error message. This is because this call is not able to generate an actual plot. The step during which the plot is actually created is when you include the geometric attribute, which, in this case is geom_point(). This is perfectly in line with the grammar of graphics since, as we have seen, the geometry represents the actual connection between the data and what is represented on the plot. This is the stage where we specify that the data should be represented as points; before that, nothing was specified about which plot we were interested in drawing.

Figure 1.7: This is an example of plotting the Orange dataset with ggplot2

In this chapter, we set up your installation of R and made sure that you are ready to start creating the ggplot2 plots. You saw the different packages available to realize plots in R and their history and relations. The graphics package is the first package that was developed in R; it represents a simple and effective tool to realize plots. Subsequently, the grid package was introduced with more advanced control of the plot elements as well as more advanced graphics functionalities. Several packages were then built on top of grid, in particular lattice and ggplot2, providing high-level functions for advanced data representation. In the next chapter, we will explore some important plot types that can be realized with ggplot2. You will also be introduced to faceting.