In jr-packages/jrGgplot2Bio: Advanced R graphics

This practical aims to guide you through some of the key ideas in ggplot2. As with the first practical, feel free to experiment. Some of the functions introduced in this practical haven't been explicitly covered in the notes. Use the built-in R help or the ggplot2 help pages at

http://had.co.nz/ggplot2/

\noindent as needed.

Over plotting

Scatter plots are very useful. However, when we have a large data set, points will be plotted on top of each other obscuring the relationship. We call this problem over plotting. There are a few techniques we can use to help, although the best solution is often problem specific.

To begin with we will create an example data frame:

## If your computer is slow when plotting reduce the value of n
library("jrGgplot2Bio")
library("ggplot2")
df = overplot_data(n = 20000)
h = ggplot(df) + geom_point(aes(x, y))

h

\noindent We can create a simple scatter plot of this data using the following command

h = ggplot(df) + geom_point(aes(x, y))

\noindent This plot isn't particularly good. Try to improve it by using a combination of:

• changing the transparency level: alpha;^[alpha takes a value between $0$ and $1$.]
• change the shape: shape=1 and shape=.'`
• use some jittering - geom_jitter.
• adding a contour to the plot using stat_density2d.
• What does

h +  stat_density2d(aes(x, y, fill = ..density..),
                contour = FALSE, geom = "tile")

do?
- What does stat_bin2d() and stat_binhex() do - add it to the plot to find out! Try varying the parameters bins and binwidth.

Displaying distributions

Lets return to the outbreaks data to look at how to visualise distributions.

data(outbreaks, package = "jrGgplot2Bio")

\noindent First lets calculate a new variable, log transformed illnesses. This is because the distribution of illnesses is very skewed, recall some of the plots from the first practical.

outbreaks$log_illness = log(outbreaks$illnesses)

\noindent We can construct a histogram of log transformed illness counts using the following commands:

ggplot(data = outbreaks) +
  geom_histogram(aes(x = log_illness))

ggplot(data = outbreaks) +
  geom_histogram(aes(x = log_illness))

\noindent to get figure 2. Let's experiment a bit.

1. Change the bins in the geom_histogram. What value do you think is best?
2. What happens when you set fill=species in the geom_histogram aesthetic? What other options can you change?^[Look at the geom_histogram help page: http://had.co.nz/ggplot2/geom_histogram.html}
3. Try geom_density. Set fill=species and change the alpha value.

Copy cat

library(dplyr)
data(yeast, package = "jrGgplot2Bio")
yeast = yeast %>% filter(class == "CYT" | class == "EXC")

The aim of this section is to recreate the graphics in figure 3. Feel free to experiment. To begin, load the package and subset the data.

library(dplyr)
data(yeast, package = "jrGgplot2Bio")
yeast = yeast %>% filter(class == "CYT" | class == "EXC")

ggplot(yeast, aes(x = gvh, y = mcg)) +
  geom_point(aes(col = class)) +
  geom_density2d(aes(col = class), alpha = 0.5) +
  xlab("von Heijne's method") +
  ylab("McGeoch's method")

1. Figure 3: Create a scatter plot of McGeoch's method for signal sequence recognition, mcg, against von Heijne's method, gvh. To get started:

ggplot(yeast, aes(x = gvh, y = mcg)) +
  geom_point() +
  xlab("von Heijne's method")

Now add the densisity lines and change the $y$-axis label. Hint: try geom_density2d and ylab(New label')`.

ggplot(yeast, aes(x = gvh, y = mcg)) +
  geom_point() +
  geom_density2d() +
  xlab("von Heijne's method") +
  ylab("McGeoch's method") 

1. Using the col aesthetic, add colour to the points and density lines dependent upon the class variable. For the final touch adjust the transparency of the density lines.

ggplot(yeast, aes(x = gvh, y = mcg)) +
  geom_point(aes(col = class)) +
  geom_density2d(aes(col = class), alpha = 0.5) +
  xlab("von Heijne's method") +
  ylab("McGeoch's method")

Solutions

Solutions are contained within this package:

vignette("solutions2", package = "jrGgplot2Bio")

jr-packages/jrGgplot2Bio documentation built on Jan. 9, 2020, 7:50 a.m.