In AndySouth/rmapteach:

ggplot2 by Hadley Wickham, is a newer alternative to the base graphics available in R. Base graphics are still useful for creating customised plots but we will not look at them here.

A) point plots with ggplot2 first go

# load the package
library("ggplot2")

# load some data on mammal sleep patterns from the package
data(msleep)

# to show the 'str'ucture of the data
str(msleep)


# first plot
ggplot(data = msleep, aes(x = bodywt, y = awake)) +
    geom_point()

# aes (for aesthetic) sets up links between the data and aspects of the plot
# + adds layers to the plot, in this case points
# geom* stands for the geometric objects you can add to a plot
# bodywt and awake are columns within the data

# add an extra column to the aes
ggplot(data = msleep, aes(x = bodywt, y = awake, col = vore)) +
    geom_point()

# if you change col= to a column that is numeric it changes the legend
ggplot(data = msleep, aes(x = bodywt, y = awake, col = sleep_rem)) +
    geom_point()

# you can store the plot as an object (in this case called p) which can make modifying it easier
p <- ggplot(data = msleep, aes(x = bodywt, y = awake, col = sleep_rem))
p <- p + geom_point()
# type the variable name to display the plot 
p

# use themes to change the overall appearance of plots
# note that when typing this in RStudio it will show you the options once you've typed th...
p + theme_bw()

A1) Exercise

Have a quick look at the online ggplot2 documentation here : http://docs.ggplot2.org/current/ Particularly look at geom_point and scroll down to the examples.

From within R you can get similar help but without the helpful plots by : ?geom_point

Try modifying a point plot with the msleep data.

B) point plots, modifying axes, label points, divide into subplots

# we can log bodywt to space the points out more
p <- ggplot(data = msleep, aes(x = log(bodywt), y = awake, col = vore)) +
     geom_point()
p

# label the points for data exploration
p <- ggplot(data = msleep, aes(x = log(bodywt), y = awake, label = name)) +
     geom_point() + 
     geom_text()
p

# divide into subplots using facet_wrap()
p <- ggplot(data = msleep, aes(x = log(bodywt), y = awake, label = name)) +
    geom_point() + 
    facet_wrap(~vore)
p

C) line plots with ggplot2

#load some time-series data from ggplot2
data(economics)

# to show the 'str'ucture of the data
str(economics)

# geom_line()
ggplot(economics, aes(date, unemploy)) + geom_line()

#note that above aes(date, unemploy) assumes that these are to set x & y respectively
# this : aes(x=date, y=unemploy) would give an identical result

# another dataset economics_long is in 'long' format
# instead of having one column for each attribute (e.g. pop and unemploy)
# it has a column (variable) which has a repeated factor values for each date
# this allows us to create a plot with multiple lines 

data(economics_long)

ggplot(economics_long, aes(date, value01, colour = variable)) +
  geom_line()

#note that for each variable the value01 column has been scaled between 0 & 1 to fit on same scale


# geom_path can be used to look at the relationship of 2 variables over time
# unemployment and savings rate
p <- ggplot(economics, aes(unemploy/pop, psavert))
p + geom_path(aes(colour = date))

# note how the unemployment rate is calculated by unemploy/pop within the aes call
# you can do other calculations between columns in there

C1) Exercise

Try using facet_wrap with the economics_long line plot to get subplots for each variable.

...

D) bigger (but still small!) data, histograms etc.

# the diamonds dataset has > 50K diamonds (rows) in it

data(diamonds)
str(diamonds)

# a scatter plot is not very informative
p <- ggplot(diamonds, aes(x = carat, y = price, col = color))
p + geom_point()

# setting the alpha (transparency) level to < 1 can improve things but not much
p <- ggplot(diamonds, aes(x = carat, y = price, col = color))
p <- p + geom_point(alpha=0.3)

# geom_smooth adds a smoothed conditional mean which can help reveal patterns
p + geom_smooth()

# geom_histogram can be used to look at distributions 
p <- ggplot(data=diamonds) + geom_histogram(aes(x=price))
p

# most of the diamonds are cheap ! (relatively)