R meetup ggplot2 + plyr presentation

gistfile1.r

# Useful online resources
# ggplot2 website: http://had.co.nz/ggplot2/
# Mailing list: https://groups.google.com/forum/?fromgroups#!forum/ggplot2
# Cookbook for R: http://wiki.stdout.org/rcookbook/


install.packages('ggplot2')
# If you already have it installed, make sure it's up-to-date
update.packages()

library(ggplot2)

# ===== Demonstrate some basic plots =====

# View the built-in mtcars dataset
mtcars

# Scatterplot
ggplot(data = mtcars, mapping = aes(x = wt, y = mpg)) + geom_point()

# Same, but with implicit arguments names
ggplot(mtcars, aes(x = wt, y = mpg)) + geom_point()

# Histogram
ggplot(mtcars, aes(x = wt)) + geom_histogram(binwidth=.25)


# View the pressure dataset
pressure 
# Points
ggplot(pressure, aes(x = temperature, y = pressure)) + geom_point()

# Points with lines (this has multiple _layers_)
ggplot(pressure, aes(x = temperature, y = pressure)) + geom_point() + geom_line()

# Bar graph (with continous x axis)
ggplot(pressure, aes(x = temperature, y = pressure)) + geom_bar(stat = "identity")
# Bar graph with discrete x axis doesn't require stat="identity"


# continuous/numeric vs. discrete/categorical/factor variables are treated very differently!
str(mtcars)
str(pressure)
str(PlantGrowth)

# View the dataset
ToothGrowth

ggplot(PlantGrowth, aes(x = group, y = weight)) + geom_point()

# Wilkinson-style dot plot
ggplot(PlantGrowth, aes(x = group, y = weight)) +
   geom_dotplot(binaxis="y", stackdir="center")

# Box plot
ggplot(PlantGrowth, aes(x = group, y = weight)) + geom_boxplot()


# ====== Setting aesthetic properties ======
# Aesthetic properties include x, y, colour, fill, shape, size
ggplot(mtcars, aes(x = wt, y = mpg)) + geom_point()
ggplot(mtcars, aes(x = wt, y = mpg)) + geom_point(colour = "red")

ggplot(mtcars, aes(x = wt, y = mpg)) + geom_point(colour = "red", size = 4)


# ====== Mapping variables to aesthetic properties ======
# Setting is different from mapping aesthetic properties

# Map disp to color
ggplot(mtcars, aes(x = wt, y = mpg, colour = disp)) + geom_point()

# Notice that you can specify mapping in the geom
# The geom will inherit other aesthetic properties from ggplot()
ggplot(mtcars, aes(x = wt, y = mpg)) + geom_point(aes(colour = disp))


# Swap some of the variables around
ggplot(mtcars, aes(x = wt, y = disp, colour = mpg)) + geom_point()

# Map disp to color -- notice that you can specify mapping in the geom
# The geom will inherit properties from ggplot()
ggplot(mtcars, aes(x = wt, y = mpg)) +
   geom_point(aes(colour = disp, shape = factor(am)))

# Treat cyl as discrete
ggplot(mtcars, aes(x = wt, y = mpg, colour = factor(cyl), shape = factor(am))) +
   geom_point()

ggplot(mtcars, aes(colour = wt, y = mpg, x = factor(cyl), shape = factor(am))) +
   geom_point()


# ====== Layers (using multiple geometric objects) =======

# Example from before
ggplot(pressure, aes(x = temperature, y = pressure)) + geom_point() + geom_line()

# Change line to red
ggplot(pressure, aes(x = temperature, y = pressure)) +
   geom_point(colour = "darkblue", size = 4) +
   geom_line(colour = "red")

# Put points on layer above line
ggplot(pressure, aes(x = temperature, y = pressure)) +
   geom_line(colour = "red") +
   geom_point(colour = "darkblue", size = 4)


# Violin plot with dot plot
ggplot(PlantGrowth, aes(x = group, y = weight)) + geom_violin() +
   geom_dotplot(binaxis="y", stackdir="center", alpha=.4)



# ====== Grouping on variables ======

# Map cyl to color
ggplot(mtcars, aes(x = wt, y = mpg, colour = cyl)) + geom_point()

# Map cyl to color, but treat is as a discrete variable
ggplot(mtcars, aes(x = wt, y = mpg, colour = factor(cyl))) + geom_point()

# Use facets with a variable
ggplot(mtcars, aes(x = wt, y = mpg)) + geom_point() +
   facet_grid(cyl ~ .)

# Faceting on two variables
ggplot(mtcars, aes(x = wt, y = mpg)) + geom_point() +
   facet_grid(cyl ~ am)


# ====== Stats (transformations on data) =======
p <- ggplot(mtcars, aes(x = wt, y = mpg)) + geom_point()

# LOESS regression
p + stat_smooth()

# Linear regression
p + stat_smooth(method = "lm")
p + stat_smooth(method = "lm", se = FALSE)  # Without confidence region


ggplot(mtcars, aes(x = wt, y = mpg, colour = factor(cyl))) + geom_point() +
   stat_smooth(method = "lm", se = FALSE)

ggplot(mtcars, aes(x = wt, y = mpg)) + geom_point() +
   stat_smooth(method = "lm", se = FALSE) +
   facet_grid(cyl ~ .)


# ====== The Grammar of Graphics =======





# ====== Using reshape2 and plyr ======
# With ggplot2, data is always in a data frame, with each column
# representing a separate variable.
# Sometimes the challenge is to get it in this format.

USJudgeRatings

# Take a subset
usj <- USJudgeRatings[1:6, 1:4]
usj

# == Convert to long format ==
library(reshape2)
melt(usj)

# We lost the names because they were row names instead of a data column
# Let's save the row names
usj
usj$name <- rownames(usj)

usj  # Notice new column

# Now it looks good
melt(usj)

# Explicitly set the id variable(s), and specify column names
melt(usj, id.vars = "name", variable.name = "measure", value.name = "number")

# Get help with ?melt.data.frame

# Now plot it
usjw <- melt(usj, id.vars = "name", variable.name = "measure", value.name = "number")
ggplot(usjw, aes(x = measure, y = number)) + geom_violin()


# == Convert to wide format ==
dcast(usjw, name ~ measure, value.var = "number")


# ====== Data transformations with plyr ======
library(plyr)

# == Summarizing by group ==

# Basic example of summarise
summarise(usjw, m = mean(number))
summarise(usjw, m = mean(number), s = sum(number))

# How to summarise each group, split on 'measure'
ddply(usjw, .(measure), summarise, m = mean(number))
ddply(usjw, .(measure), summarise, m = mean(number), n = length(number))

# Summarizing splitting on 'name' and 'measure' (but it's not very interesting)
ddply(usjw, .(name, measure), summarise, m = mean(number), n = length(number))


# == Transforming by group ==

# Basic example of transform
transform(usjw, minus5 = number - 5)
transform(usjw, m = mean(number))
transform(usjw, normalized = number - mean(number))

# How to transform each group, split
ddply(usjw, .(measure), transform, m = mean(number), normalized = number - mean(number))
ddply(usjw, .(measure), transform, m = mean(number), normalized = number - mean(number))