In jr-packages/jrGgplot2: Jumping Rivers: Advanced Graphics in R

This practical aims to guide you through some of the key ideas in data manipulation. I've tried to construct this practical in such a way that you get to experiment with the various tools. Feel free to experiment!

library("ggplot2")
data(aphids, package = "jrGgplot2")

Factors

When using ggplot2, the easiest way of rearranging the graph or to alter labels is to manipulate the data set. Consider the mpg data set:

data(mpg, package = "ggplot2")

Suppose we generate a scatter plot of engine displacement against highway mpg.

g = ggplot(data = mpg, aes(x = displ, y = hwy)) +
  geom_point()

\noindent Next, we add a loess line, conditional on the drive type:

g + stat_smooth(aes(colour = drv))

\noindent While this graph is suitable for exploring the data; for publication, we would like to rename the axis and legend labels. To change the axis labels, we can rename the data frame columns or use xlab and ylab. To change the order of the legend, we need to manipulate the data. Since drv is a character, we could use:

mpg[mpg$drv == "4", ]$drv = "4wd"
mpg[mpg$drv == "f", ]$drv = "Front"
mpg[mpg$drv == "r", ]$drv = "Rear"

\noindent However the legend will still be ordered alphabetically. Instead, we can use a factor:

##Reload the data just to make sure
data(mpg, package = "ggplot2")
mpg$drv = factor(mpg$drv, labels = c("4wd", "Front", "Rear"))

\noindent To change the order of the, we need to use the factor function:

mpg$drv = factor(mpg$drv,
                 levels = c("Front", "Rear", "4wd"))

\noindent The legend now displays the labels in the order: Front, Rear and 4wd.

aphids$Block = factor(aphids$Block)
aphids$Water = factor(aphids$Water,
                      levels = c("Low", "Medium", "High"))
ga = ggplot(data = aphids) +
  geom_point(aes(Time, Aphids, colour = Block)) +
  facet_grid(Nitrogen ~ Water) +
  geom_line(aes(Time, Aphids, colour = Block)) +
  theme_bw()

print(ga)

Aphids

This data set consists of seven observations on cotton aphid counts on twenty randomly chosen leaves in each plot, for twenty-seven treatment-block combinations. The data were recorded in July 2004 in Lamesa, Texas. The treatments consisted of three nitrogen levels (block, variable and none), three irrigation levels (low, medium and high) and three blocks, each being a distinct area. Irrigation treatments were randomly assigned within each block as whole plots. Nitrogen treatments were randomly assigned within each whole block as split plots.

data(aphids, package = "jrGgplot2")

\noindent The sampling times are once per week.

\newthought{Reproduce} figure 1. Here are some hints to get you started. The key idea is to think of the plot in terms of layers. So

• Leave the ordering of factors to the end
• The plot contains a combination of geom_line and geom_point.
• You can change the x-axis label using

 + xlab("Time")

• Change the theme using theme_bw()

\newpage

##Code for figure 1
aphids$Block = factor(aphids$Block)
aphids$Water = factor(aphids$Water,
                      levels = c("Low", "Medium", "High"))
ga = ggplot(data = aphids) +
  geom_point(aes(Time, Aphids, colour = Block)) +
  facet_grid(Nitrogen ~ Water) +
  geom_line(aes(Time, Aphids, colour = Block)) +
  theme_bw()

The Beauty data set

First load the beauty data set

data(Beauty, package = "jrGgplot2")

\noindent In practical 1, we split data up by both gender and age:

Beauty$dec = signif(Beauty$age, 1)
g = ggplot(data = Beauty)
g1 = g + geom_bar(aes(x = gender, fill = factor(dec)))

g1

\noindent to get figure 2. Rather than using the fill aesthetic, redo the plot but use facet_grid and facet_wrap. For example,

g2 = g + geom_bar(aes(x = gender)) +
  facet_grid(. ~ dec)

\noindent Experiment with:

• the margins argument

g + geom_bar(aes(x = gender)) + facet_grid(. ~ dec, margins = TRUE)

• the scales=free_y'` argument

## Notice that the females have disappeared from the "70" facet.
## Probably not what we wanted.
g + geom_bar(aes(x = gender)) + facet_grid(. ~ dec, scales = "free_x")

• the layout, i.e. column or row.

g + geom_bar(aes(x = gender)) + facet_grid(dec ~ .)

How would you change the panel labels?

##Relabel the factor
Beauty$dec = factor(Beauty$dec,
                    labels = c("Thirties", "Forties", "Fifties", "Sixties", "Seventies"))

## Plot as before
ggplot(data = Beauty) +
  geom_bar(aes(x = gender)) +
  facet_grid(dec ~ .)

The Google data set

data(google, package = "jrGgplot2")
g = ggplot(google) +
  geom_point(aes(Rank, Users), alpha = 0.2) +
  scale_y_log10(limit = c(1e0, 1e9))  +
  facet_grid(Advertising ~ .) +
  geom_point(data = subset(google, Category == "Social Networks"),
             aes(Rank, Users), colour = "Red",  size = 2)

Google recently released a data set of the top 1000 websites. The data set contains the following categories: Rank, Site, Category, Users, Views and Advertising. First we load the data

data(google, package = "jrGgplot2")

g

1. Create a scatter of plot of Rank and Views.
2. Using scale_y_log10() transform the y scale.
3. Uses facets to split the plot by its advertising status.
4. Use another geom_point() layer to highlight the Social Networks sites to figure 3.

g = ggplot(google) +
  geom_point(aes(Rank, Users), alpha = 0.2) +
  scale_y_log10(limit = c(1e0, 1e9))  +
  facet_grid(Advertising ~ .) +
  geom_point(data = subset(google, Category == "Social Networks"),
             aes(Rank, Users), colour = "Red",  size = 2)

An example data set

An experiment was conducted: two treatments, A and B, were tested. The data can be downloaded into R using the following commands:

data(cell_data, package = "jrGgplot2")

\noindent Within each treatment group, there are two patient types: Case and Control. What plot do you think would be most suitable for this data?

First we'll create a base object

##This doesn't plot anything
g = ggplot(cell_data, aes(treatment, values)) +
  facet_grid(.~type)

Experiment plotting the data using boxplots, jittered points, histograms, errors, etc. Which methods is optimal (for this data set).

##Boxplot
g + geom_boxplot()

## Dotplot
g + geom_dotplot(binaxis = "y", stackdir = "center",
                 colour = "blue", fill = "blue")

## Plain (jittered) points
g + geom_jitter()

## Dotplots + error bars
g +  geom_dotplot(binaxis = "y", stackdir = "center",
                  colour = "blue", fill = "blue") +
  stat_summary(geom = "errorbar",
               fun.ymin = min, fun.ymax = max, fun.y = mean,
               width = 0.2)

Solutions

Solutions are contained within this package:

library("jrGgplot2")
vignette("solutions3", package = "jrGgplot2")

jr-packages/jrGgplot2 documentation built on Sept. 20, 2020, 2:59 a.m.