Data Visualization Using R

Introduction

• ggplot2 is a powerful R package that produces data visualizations easily and intuitively.

• Since this package is not built into the language, you will need to install it by running in the console the following code:

install.packages("ggplot2")

• For each of your R sessions, you need to load the library before you use it, which can be done as follows:

library("ggplot2")

• Once this is completed, you can now proceed with using this package. For demonstration purposes, I will use the diamonds dataset that is built into the package for revealing the packages functionalities. In addition, I will provide examples of two other datasets pre-installed with R, which are the iris and mtcars datasets.

• The dataset can be accessed with the data function:

data(diamonds)
data(iris)
data(mtcars)

• One of the ways that we can view what the dataset contains is by using the View() function:

View(diamonds)
View(iris)
View(mtcars)

• Furthermore, R’s help() function can give us a description of the diamonds dataset and details about each of the columns.

help(diamonds)
help(iris)
help(mtcars)

• For the remainder

Introduction to ggplot2

• To start getting familiar with ggplot2, lets attempt to understand the relationship between the dataset diamonds’s attributes through the use of a scatter plot.

• Before we do so, we need to set up the graph’s “aesthetics”, which is the dimension of a graph that we can perceive visually such as its x- and y- axes, color, size, and shape.

• Lets create a ggplot object where the attribute caret is on the x-axis and price is on the y-axis:

d <- ggplot(data = diamonds, aes(x=carat, y=price)) 
d + geom_point()

i <- ggplot(data=iris, aes(x=Sepal.Length, y=Sepal.Width)) 
i + geom_point()

m <- ggplot(data = mtcars, aes(x = wt, y = mpg)) 
m + geom_point()

• As can be observed, this graph consists of 3 parts:

1. The data being graphed (diamonds dataset)

2. Mapping of aesthetics to the attributes with the aes() function and assignments “x=carat” (carat on the x-axis) and “y=price” (price on the y-axis)

3. Lastly, the assigned layer, which defines what type of graph it is. Since we want a scatter plot, the corresponding layer is geom_pont.

• Nonetheless, we can still explore other attributes of the data that might tell a story themselves such as how the clarity affects the price:

d + geom_point(aes(colour=clarity))

i + geom_point(aes(colour=Species))

m + geom_point(aes(colour=factor(cyl)))

• How about the effect of both clarity and cut?

d + geom_point(aes(colour=clarity, size=cut))

## Warning: Using size for a discrete variable is not advised.

m + geom_point(aes(colour=factor(cyl), size = qsec))

• We could use the shape to represent the cut:

d + geom_point(aes(colour=clarity, shape=cut))

i + geom_point(aes(colour=Species, shape=Species))

m + geom_point(aes(colour = factor(cyl), shape = factor(cyl)))

• Up to now, we have been working with one layer, but we can add more by using the plus sign. To observe this, lets use the geom_smooth layer, which adds a smoothing curve that shows the general trend of the data:

d + geom_point() + geom_smooth()

## `geom_smooth()` using method = 'gam'

i + geom_point(aes(colour = Species)) + geom_smooth(method = "lm")

• The gray area around the curve is a confidence interval, suggesting how much uncertainty there is in this smoothing curve. If we want to turn off the confidence interval, we can add an option to the geom_smooth later; specifically “se=FALSE”, where “s.e.” stands for “standard error.

d + geom_point() + geom_smooth(se=FALSE)

## `geom_smooth()` using method = 'gam'

m + geom_point(aes(colour = factor(cyl))) + geom_smooth(method="lm", se=FALSE)

Faceting and Additional Options

• “Faceting” is another way of communicating information by dividing your plot into multiple plots, one for each level, letting you view them separately. This is done in ggplot with the facet_wrap() function where you put a tilde (~) along with the attribute you would like to divide the plots by.

d + geom_point(aes(colour=cut)) + facet_wrap(~ clarity)

i + geom_point() + facet_grid(. ~ Species) + geom_smooth(method = "lm")

• You can even divide your graph based on two different attributes, such as both color and clarity, using facetgrid(). In this case that would be “facetgrid(”, then you put “color ~ clarity”, where the tilde (~) means “explained by.”

d + geom_point(aes(colour=cut)) + facet_grid(color ~ clarity)

ggplot(data=mtcars, aes(x=mpg, y=disp)) + 
  geom_point(aes(color = carb)) + 
  facet_grid(cyl ~ gear)

• You might want to set a title, or set the x or y axis labels manually. You change these options by adding to the end of the line of code. To set the title, you would use the ggtitle function:

d + geom_point() + ggtitle("My scatter plot")

• To change the x- or y- axis labels, you would add “xlab” for “x label”, then your custom label:

ggplot(diamonds, aes(x=carat, y=price)) +
  geom_point() +
  ggtitle("My scatter plot") +
  xlab("Weight (carats)")

• You might also want to limit the range of the x or the y axes. You can do this with the xlim or ylim options:

ggplot(diamonds, aes(x=carat, y=price)) +
  geom_point() +
  ggtitle("My scatter plot") +
  xlab("Weight (carats)") +
  xlim(0, 2)

## Warning: Removed 1889 rows containing missing values (geom_point).

ggplot(diamonds, aes(x=carat, y=price)) +
  geom_point() +
  ggtitle("My scatter plot") +
  xlab("Weight (carats)") +
  ylim(0, 10000)

## Warning: Removed 5222 rows containing missing values (geom_point).

• Another possibility is to put one of the axes on a log scale. You can do this with the scale_y_log10() function:

ggplot(diamonds, aes(x=carat, y=price)) +
  geom_point() +
  ggtitle("My scatter plot") +
  xlab("Weight (carats)") +
  scale_y_log10() 

• There are many other available options and customizations: each gets added to the end of the plot just like these.

Line Graphs

• For line graphs, the data points must be grouped so that it knows which points to connect.

m + geom_line(aes(colour = as.factor(cyl)))

Histograms and Density Plots

• To look at just one dimension of your data and observe its distribution, we can use a histogram, and all you need to do from before is to change your layer from geom_point() to geom_histogram():

ggplot(diamonds, aes(x=price)) +
  geom_histogram()

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

• You can change the width of each bin as an option to the geom_histogram layer. You can make them wider or thinner:

ggplot(diamonds, aes(x=price)) +
  geom_histogram(binwidth=2000)

ggplot(data = mtcars, aes(x = mpg, fill=cyl)) +
  geom_histogram(binwidth = 1)

• Let’s say you want to add another attribute to this histogram to see its effect on the density: for example, to make a stacked histogram based on the clarity attribute. Try adding the “fill” aesthetic:

ggplot(diamonds, aes(x=price, fill=clarity)) +
  geom_histogram()

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

• Another way to view the distribution is as a density curve. You can do this by changing geom_histogram to geom_density. Remove the fill attribute.

ggplot(diamonds, aes(x=price)) + 
  geom_density()

• If you want to divide this density curve up based on one of your attributes, you can use the color aesthetic instead of fill. For example, you can add color=cut.

ggplot(diamonds, aes(x=price, color=cut)) +
  geom_density()

ggplot(iris, aes(x = Sepal.Length, color = Species, fill=Species)) +
  geom_density(alpha=0.3)

Boxplots

• One common method in statistics for comparing multiple densities is to use a box plot. A box plot has two attributes: an x, which is usually a classification into categories, and y, the actual variable that you’re comparing.

ggplot(diamonds, aes(x=color, y=price)) +
  geom_boxplot()

ggplot(iris, aes(x=Species,y=Sepal.Length)) + 
  geom_boxplot()

• The median is the thick black line, the edges of the box show the 25th and 75th quantiles of the data respectively, and these points are outliers that lie far outside the expected range of the data.

Output: Saving Your Plots

• Let’s create a scatter plot based on our diamonds data, but save it to a variable, called p.

p = ggplot(diamonds, aes(x=carat, y=price)) +
  geom_point()

• Now we can save that plot to a file, instead of displaying it in our window, by using ggsave:

1. PNG

ggsave(filename="diamonds.png", p)

2. PDF

ggsave(filename="diamonds.pdf", p)

3. JPEG

ggsave(filename="diamonds.jpeg", p)

• One useful shortcut is that if you just displayed a plot, ggsave will know to save that plot by default when you perform ggsave. You don’t even have to tell it which plot you’re saving.

ggplot(diamonds, aes(x=carat, y=price)) +
  geom_point()

ggsave("diamonds.png")