Introduction to R Graphics with ggplot2
Table of Contents
- Introduction
- Geometric Objects And Aesthetics
- Statistical Transformations
- Scales
- Faceting
- Themes
- The #1 FAQ
- Putting It All Together
- Challenge Solution prototype
Introduction
Materials and setup
Everyone: Download workshop materials:
- Download materials from http://www.math-evry.cnrs.fr/_media/members/aguilloux/enseignements/m2isg/rgraphics_site.zip
- Extract the zip file containing the materials to your desktop
Workshop notes are available in .hmtl format. Open a file browser, navigate to your desktop and open Rgraphics.html
Starting A The End
My goal: by the end of the workshop you will be able to reproduce this graphic from the Economist:
Why ggplot2?
Advantages of ggplot2
- consistent underlying
grammar of graphics(Wilkinson, 2005) - plot specification at a high level of abstraction
- very flexible
- theme system for polishing plot appearance
- mature and complete graphics system
- many users, active mailing list
That said, there are some things you cannot (or should not) do With ggplot2:
- 3-dimensional graphics (see the rgl package)
- Graph-theory type graphs (nodes/edges layout; see the igraph package)
- Interactive graphics (see the ggvis package)
What Is The Grammar Of Graphics?
The basic idea: independently specify plot building blocks and combine them to create just about any kind of graphical display you want. Building blocks of a graph include:
- data
- aesthetic mapping
- geometric object
- statistical transformations
- scales
- coordinate system
- position adjustments
- faceting
Example Data: Housing prices
Let's look at housing prices.
housing <- read.csv("dataSets/landdata-states.csv") head(housing[1:5])
> housing <- read.csv("dataSets/landdata-states.csv")
> head(housing[1:5])
State region Date Home.Value Structure.Cost
1 AK West 20101 224952 160599
2 AK West 20102 225511 160252
3 AK West 20093 225820 163791
4 AK West 20094 224994 161787
5 AK West 20074 234590 155400
6 AK West 20081 233714 157458
>
(Data from https://www.lincolninst.edu/subcenters/land-values/land-prices-by-state.asp)
housing <- read.csv("dataSets/landdata-states.csv") housing$Year <- as.numeric(substr(housing$Date, 1, 4)) housing$Qrtr <- as.numeric(substr(housing$Date, 5, 5)) housing$Date <- housing$Year + housing$Qrtr/4
ggplot2 VS Base Graphics
Compared to base graphics, ggplot2
- is more verbose for simple / canned graphics
- is less verbose for complex / custom graphics
- does not have methods (data should always be in a
data.frame) - uses a different system for adding plot elements
ggplot2 VS Base for simple graphs
Base graphics histogram example:
hist(housing$Home.Value)
ggplot2 histogram example:
library(ggplot2)
ggplot(housing, aes(x = Home.Value)) +
geom_histogram()
Base wins!
ggplot2 Base graphics VS ggplot for more complex graphs:
Base colored scatter plot example:
plot(Home.Value ~ Date,
data=subset(housing, State == "MA"))
points(Home.Value ~ Date, col="red",
data=subset(housing, State == "TX"))
legend(19750, 400000,
c("MA", "TX"), title="State",
col=c("black", "red"),
pch=c(1, 1))
ggplot2 colored scatter plot example:
ggplot(subset(housing, State %in% c("MA", "TX")), aes(x=Date, y=Home.Value, color=State))+ geom_point()
ggplot2 wins!
Geometric Objects And Aesthetics
Aesthetic Mapping
In ggplot land aesthetic means "something you can see". Examples include:
- position (i.e., on the x and y axes)
- color ("outside" color)
- fill ("inside" color)
- shape (of points)
- linetype
- size
Each type of geom accepts only a subset of all aesthetics–refer to the geom help pages to see what mappings each geom accepts. Aesthetic mappings are set with the aes() function.
Geometic Objects (geom)
Geometric objects are the actual marks we put on a plot. Examples include:
- points (
geom_point, for scatter plots, dot plots, etc) - lines (
geom_line, for time series, trend lines, etc) - boxplot (
geom_boxplot, for, well, boxplots!)
A plot must have at least one geom; there is no upper limit. You can add a geom to a plot using the + operator
You can get a list of available geometric objects using the code below:
help.search("geom_", package = "ggplot2")
or simply type geom_<tab> in any good R IDE (such as Rstudio or ESS) to see a list of functions starting with geom_.
Points (Scatterplot)
Now that we know about geometric objects and aesthetic mapping, we can make a ggplot. geom_point requires mappings for x and y, all others are optional.
hp2001Q1 <- subset(housing, Date == 2001.25) ggplot(hp2001Q1, aes(y = Structure.Cost, x = Land.Value)) + geom_point()
ggplot(hp2001Q1,
aes(y = Structure.Cost, x = log(Land.Value))) +
geom_point()
Lines (Prediction Line)
A plot constructed with ggplot can have more than one geom. In that case the mappings established in the ggplot() call are plot defaults that can be added to or overridden. Our plot could use a regression line:
hp2001Q1$pred.SC <- predict(lm(Structure.Cost ~ log(Land.Value), data = hp2001Q1)) p1 <- ggplot(hp2001Q1, aes(x = log(Land.Value), y = Structure.Cost)) p1 + geom_point(aes(color = Home.Value)) + geom_line(aes(y = pred.SC))
Smoothers
Not all geometric objects are simple shapes–the smooth geom includes a line and a ribbon.
p1 + geom_point(aes(color = Home.Value)) + geom_smooth()
Text (Label Points)
Each geom accepts a particualar set of mappings–for example geom_text() accepts a labels mapping.
p1 +
geom_text(aes(label=State), size = 3)
## install.packages("ggrepel") library("ggrepel") p1 + geom_point() + geom_text_repel(aes(label=State), size = 3)
Aesthetic Mapping VS Assignment
Note that variables are mapped to aesthetics with the aes() function, while fixed aesthetics are set outside the aes() call. This sometimes leads to confusion, as in this example:
p1 + geom_point(aes(size = 2),# incorrect! 2 is not a variable color="red") # this is fine -- all points red
Mapping Variables To Other Aesthetics
Other aesthetics are mapped in the same way as x and y in the previous example.
p1 + geom_point(aes(color=Home.Value, shape = region))
Exercise I
The data for the exercises is available in the dataSets/EconomistData.csv file. Read it in with
dat <- read.csv("dataSets/EconomistData.csv") head(dat) ggplot(dat, aes(x = CPI, y = HDI, size = HDI.Rank)) + geom_point()
> dat <- read.csv("dataSets/EconomistData.csv")
>
Original sources for these data are http://www.transparency.org/content/download/64476/1031428 http://hdrstats.undp.org/en/indicators/display_cf_xls_indicator.cfm?indicator_id=103106&lang=en
These data consist of Human Development Index and Corruption Perception Index scores for several countries.
- Create a scatter plot with CPI on the x axis and HDI on the y axis.
- Color the points blue.
- Map the color of the the points to Region.
- Make the points bigger by setting size to 2
- Map the size of the points to HDI.Rank
Statistical Transformations
Statistical Transformations
Some plot types (such as scatterplots) do not require transformations–each point is plotted at x and y coordinates equal to the original value. Other plots, such as boxplots, histograms, prediction lines etc. require statistical transformations:
- for a boxplot the y values must be transformed to the median and 1.5(IQR)
- for a smoother smother the y values must be transformed into predicted values
Each geom has a default statistic, but these can be changed. For example, the default statistic for geom_bar is stat_count:
args(geom_histogram) args(stat_bin)
> args(geom_histogram)
function (mapping = NULL, data = NULL, stat = "bin", binwidth = NULL,
bins = NULL, origin = NULL, right = FALSE, position = "stack",
na.rm = FALSE, show.legend = NA, inherit.aes = TRUE, ...)
NULL
> args(stat_bin)
function (mapping = NULL, data = NULL, geom = "bar", position = "stack",
width = 0.9, drop = FALSE, right = FALSE, binwidth = NULL,
bins = NULL, origin = NULL, breaks = NULL, na.rm = FALSE,
show.legend = NA, inherit.aes = TRUE, ...)
NULL
>
Setting Statistical Transformation Arguments
Arguments to stat_ functions can be passed through geom_ functions. This can be slightly annoying because in order to change it you have to first determine which stat the geom uses, then determine the arguments to that stat.
For example, here is the default histogram of Home.Value:
p2 <- ggplot(housing, aes(x = Home.Value))
p2 + geom_histogram()
The binwidth looks reasonable by default, but we can change it by passing the binwidth argument to the stat_bin function:
p2 + geom_histogram(stat = "bin", binwidth=4000)
Changing The Statistical Transformation
Sometimes the default statistical transformation is not what you need. This is often the case with pre-summarized data:
housing.sum <- aggregate(housing["Home.Value"], housing["State"], FUN=mean) rbind(head(housing.sum), tail(housing.sum))
> housing.sum <- aggregate(housing["Home.Value"], housing["State"], FUN=mean) > rbind(head(housing.sum), tail(housing.sum)) State Home.Value 1 AK 147385 2 AL 92545 3 AR 82077 4 AZ 140756 5 CA 282808 6 CO 158176 46 VA 155391 47 VT 132395 48 WA 178523 49 WI 108359 50 WV 77162 51 WY 122897 >
ggplot(housing.sum, aes(x=State, y=Home.Value)) + geom_bar()
ggplot(housing.sum, aes(x=State, y=Home.Value)) + geom_bar() Error: stat_count() must not be used with a y aesthetic.
What is the problem with the previous plot? Basically we take binned and summarized data and ask ggplot to bin and summarize it again (remember, geom_bar defaults to stat = stat_count); obviously this will not work. We can fix it by telling geom_bar to use a different statistical transformation function:
ggplot(housing.sum, aes(x=State, y=Home.Value)) +
geom_bar(stat="identity")
Exercise II
- Re-create a scatter plot with CPI on the x axis and HDI on the y axis (as you did in the previous exercise).
- Overlay a smoothing line on top of the scatter plot using geomsmooth.
- Overlay a smoothing line on top of the scatter plot using geomsmooth, but use a linear model for the predictions. Hint: see
?stat_smooth. - Overlay a smoothling line on top of the scatter plot using geomline. Hint: change the statistical transformation.
- BONUS: Overlay a smoothing line on top of the scatter plot using the default loess method, but make it less smooth. Hint: see
?loess.
Scales
Scales: Controlling Aesthetic Mapping
Aesthetic mapping (i.e., with aes()) only says that a variable should be mapped to an aesthetic. It doesn't say how that should happen. For example, when mapping a variable to shape with aes(shape = x) you don't say what shapes should be used. Similarly, aes(color = z) doesn't say what colors should be used. Describing what colors/shapes/sizes etc. to use is done by modifying the corresponding scale. In ggplot2 scales include
- position
- color and fill
- size
- shape
- line type
Scales are modified with a series of functions using a scale_<aesthetic>_<type> naming scheme. Try typing scale_<tab> to see a list of scale modification functions.
Common Scale Arguments
The following arguments are common to most scales in ggplot2:
- name
- the first argument gives the axis or legend title
- limits
- the minimum and maximum of the scale
- breaks
- the points along the scale where labels should appear
- labels
- the labels that appear at each break
Specific scale functions may have additional arguments; for example, the scale_color_continuous function has arguments low and high for setting the colors at the low and high end of the scale.
Scale Modification Examples
Start by constructing a dotplot showing the distribution of home values by Date and State.
p3 <- ggplot(housing, aes(x = State, y = Home.Price.Index)) + theme(legend.position="top", axis.text=element_text(size = 6)) (p4 <- p3 + geom_point(aes(color = Date), alpha = 0.5, size = 1.5, position = position_jitter(width = 0.25, height = 0)))
Now modify the breaks for the x axis and color scales
p4 + scale_x_discrete(name="State Abbreviation") + scale_color_continuous(name="", breaks = c(1976, 1994, 2013), labels = c("'76", "'94", "'13"))
Next change the low and high values to blue and red:
p4 + scale_x_discrete(name="State Abbreviation") + scale_color_continuous(name="", breaks = c(1976, 1994, 2013), labels = c("'76", "'94", "'13"), low = "blue", high = "red")
p4 + scale_color_continuous(name="", breaks = c(1976, 1994, 2013), labels = c("'76", "'94", "'13"), low = muted("blue"), high = muted("red"))
Using different color scales
ggplot2 has a wide variety of color scales; here is an example using scale_color_gradient2 to interpolate between three different colors.
p4 + scale_color_gradient2(name="", breaks = c(1976, 1994, 2013), labels = c("'76", "'94", "'13"), low = muted("blue"), high = muted("red"), mid = "gray60", midpoint = 1994)
Available Scales
- Partial combination matrix of available scales
| Scale | Types | Examples |
|---|---|---|
| scalecolor_ | identity | scalefillcontinuous |
| scalefill_ | manual | scalecolordiscrete |
| scalesize_ | continuous | scalesizemanual |
| discrete | scalesizediscrete | |
| scaleshape_ | discrete | scaleshapediscrete |
| scalelinetype_ | identity | scaleshapemanual |
| manual | scalelinetypediscrete | |
| scalex_ | continuous | scalexcontinuous |
| scaley_ | discrete | scaleydiscrete |
| reverse | scalexlog | |
| log | scaleyreverse | |
| date | scalexdate | |
| datetime | scaleydatetime |
Note that in RStudio you can type scale_ followed by TAB to get the whole list of available scales.
Exercise III
- Create a scatter plot with CPI on the x axis and HDI on the y axis. Color the points to indicate region.
- Modify the x, y, and color scales so that they have more easily-understood names (e.g., spell out "Human development Index" instead of "HDI").
- Modify the color scale to use specific values of your choosing. Hint: see
?scale_color_manual.
Faceting
Faceting
- Faceting is
ggplot2parlance for small multiples - The idea is to create separate graphs for subsets of data
ggplot2offers two functions for creating small multiples:facet_wrap(): define subsets as the levels of a single grouping variablefacet_grid(): define subsets as the crossing of two grouping variables
- Facilitates comparison among plots, not just of geoms within a plot
What is the trend in housing prices in each state?
- Start by using a technique we already know–map State to color:
p5 <- ggplot(housing, aes(x = Date, y = Home.Value))
p5 + geom_line(aes(color = State))
There are two problems here–there are too many states to distinguish each one by color, and the lines obscure one another.
Faceting to the rescue
We can remedy the deficiencies of the previous plot by faceting by state rather than mapping state to color.
(p5 <- p5 + geom_line() + facet_wrap(~State, ncol = 10))
There is also a facet_grid() function for faceting in two dimensions.
Themes
Themes
The ggplot2 theme system handles non-data plot elements such as
- Axis labels
- Plot background
- Facet label backround
- Legend appearance
Built-in themes include:
theme_gray()(default)theme_bw()theme_classc()
p5 + theme_linedraw()
p5 + theme_light()
Overriding theme defaults
Specific theme elements can be overridden using theme(). For example:
p5 + theme_minimal() +
theme(text = element_text(color = "turquoise"))
All theme options are documented in ?theme.
Creating and saving new themes
You can create new themes, as in the following example:
theme_new <- theme_bw() + theme(plot.background = element_rect(size = 1, color = "blue", fill = "black"), text=element_text(size = 12, family = "Serif", color = "ivory"), axis.text.y = element_text(colour = "purple"), axis.text.x = element_text(colour = "red"), panel.background = element_rect(fill = "pink"), strip.background = element_rect(fill = muted("orange"))) p5 + theme_new
The #1 FAQ
Map Aesthetic To Different Columns
The most frequently asked question goes something like this: I have two variables in my data.frame, and I'd like to plot them as separate points, with different color depending on which variable it is. How do I do that?
Wrong
housing.byyear <- aggregate(cbind(Home.Value, Land.Value) ~ Date, data = housing, mean) ggplot(housing.byyear, aes(x=Date)) + geom_line(aes(y=Home.Value), color="red") + geom_line(aes(y=Land.Value), color="blue") #
Right
library(tidyr) home.land.byyear <- gather(housing.byyear, value = "value", key = "type", Home.Value, Land.Value) ggplot(home.land.byyear, aes(x=Date, y=value, color=type)) + geom_line()
Putting It All Together
Challenge: Recreate This Economist Graph
Graph source: http://www.economist.com/node/21541178
Building off of the graphics you created in the previous exercises, put the finishing touches to make it as close as possible to the original economist graph.
Challenge Solution prototype
Lets start by creating the basic scatter plot, then we can make a list of things that need to be added or changed. The basic plot loogs like this:
dat <- read.csv("dataSets/EconomistData.csv") pc1 <- ggplot(dat, aes(x = CPI, y = HDI, color = Region)) pc1 + geom_point()
To complete this graph we need to:
[ ]add a trend line[ ]change the point shape to open circle[ ]change the order and labels of Region[ ]label select points[ ]fix up the tick marks and labels[ ]move color legend to the top[ ]title, label axes, remove legend title[ ]theme the graph with no vertical guides[ ]add model R2 (hard)[ ]add sources note (hard)[ ]final touches to make it perfect (use image editor for this)
Adding the trend line
Adding the trend line is not too difficult, though we need to guess at the model being displyed on the graph. A little bit of trial and error leds to
(pc2 <- pc1 + geom_smooth(aes(group = 1), method = "lm", formula = y ~ log(x), se = FALSE, color = "red")) + geom_point()
Notice that we put the geom_line layer first so that it will be plotted underneath the points, as was done on the original graph.
Use open points
This one is a little tricky. We know that we can change the shape with the shape argument, what what value do we set shape to? The example shown in ?shape can help us:
## A look at all 25 symbols df2 <- data.frame(x = 1:5 , y = 1:25, z = 1:25) s <- ggplot(df2, aes(x = x, y = y)) s + geom_point(aes(shape = z), size = 4) + scale_shape_identity() ## While all symbols have a foreground colour, symbols 19-25 also take a ## background colour (fill) s + geom_point(aes(shape = z), size = 4, colour = "Red") + scale_shape_identity() s + geom_point(aes(shape = z), size = 4, colour = "Red", fill = "Black") + scale_shape_identity()
This shows us that shape 1 is an open circle, so
pc2 + geom_point(shape = 1, size = 4)
That is better, but unfortunately the size of the line around the points is much narrower than on the original.
(pc3 <- pc2 + geom_point(shape = 1, size = 2.5, stroke = 1.25))
Labelling points
This one is tricky in a couple of ways. First, there is no attribute in the data that separates points that should be labelled from points that should not be. So the first step is to identify those points.
pointsToLabel <- c("Russia", "Venezuela", "Iraq", "Myanmar", "Sudan", "Afghanistan", "Congo", "Greece", "Argentina", "Brazil", "India", "Italy", "China", "South Africa", "Spane", "Botswana", "Cape Verde", "Bhutan", "Rwanda", "France", "United States", "Germany", "Britain", "Barbados", "Norway", "Japan", "New Zealand", "Singapore")
Now we can label these points using geom_text, like this:
#+ENDSRC
(pc4 <- pc3 + geom_text(aes(label = Country), color = "gray20", data = subset(dat, Country %in% pointsToLabel)))
This more or less gets the information across, but the labels overlap in a most unpleasing fashion. We can use the ggrepel package to make things better, but if you want perfection you will probably have to do some hand-adjustment.
library("ggrepel") (pc4 <- pc3 + geom_text_repel(aes(label = Country), color = "gray20", data = subset(dat, Country %in% pointsToLabel), force = 10))
Change the region labels and order
Thinkgs are starting to come together. There are just a couple more things we need to add, and then all that will be left are themeing changes.
Comparing our graph to the original we notice that the labels and order of the Regions in the color legend differ. To correct this we need to change both the labels and order of the Region variable. We can do this with the factor function.
dat$Region <- factor(dat$Region, levels = c("EU W. Europe", "Americas", "Asia Pacific", "East EU Cemt Asia", "MENA", "SSA"), labels = c("OECD", "Americas", "Asia &\nOceania", "Central &\nEastern Europe", "Middle East &\nnorth Africa", "Sub-Saharan\nAfrica"))
Now when we construct the plot using these data the order should appear as it does in the original.
pc4$data <- dat
pc4
Add title and format axes
The next step is to add the title and format the axes. We do that using the scales system in ggplot2.
library(grid) (pc5 <- pc4 + scale_x_continuous(name = "Corruption Perceptions Index, 2011 (10=least corrupt)", limits = c(.9, 10.5), breaks = 1:10) + scale_y_continuous(name = "Human Development Index, 2011 (1=Best)", limits = c(0.2, 1.0), breaks = seq(0.2, 1.0, by = 0.1)) + scale_color_manual(name = "", values = c("#24576D", "#099DD7", "#28AADC", "#248E84", "#F2583F", "#96503F")) + ggtitle("Corruption and Human development"))
Theme tweaks
Our graph is almost there. To finish up, we need to adjust some of the theme elements, and label the axes and legends. This part usually involves some trial and error as you figure out where things need to be positioned. To see what these various theme settings do you can change them and observe the results.
library(grid) # for the 'unit' function (pc6 <- pc5 + theme_minimal() + # start with a minimal theme and add what we need theme(text = element_text(color = "gray20"), legend.position = c("top"), # position the legend in the upper left legend.direction = "horizontal", legend.justification = 0.1, # anchor point for legend.position. legend.text = element_text(size = 11, color = "gray10"), axis.text = element_text(face = "italic"), axis.title.x = element_text(vjust = -1), # move title away from axis axis.title.y = element_text(vjust = 2), # move away for axis axis.ticks.y = element_blank(), # element_blank() is how we remove elements axis.line = element_line(color = "gray40", size = 0.5), axis.line.y = element_blank(), panel.grid.major = element_line(color = "gray50", size = 0.5), panel.grid.major.x = element_blank() ))
Add model R2 and source note
The last bit of information that we want to have on the graph is the variance explained by the model represented by the trend line. Lets fit that model and pull out the R2 first, then think about how to get it onto the graph.
(mR2 <- summary(lm(HDI ~ log(CPI), data = dat))$r.squared)
> (mR2 <- summary(lm(HDI ~ log(CPI), data = dat))$r.squared) [1] 0.5212859 >
OK, now that we've calculated the values, let's think about how to get them on the graph. ggplot2 has an annotate function, but this is not convenient for adding elements outside the plot area. The grid package has nice functions for doing this, so we'll use those.
And here it is, our final version!
library(grid) png(file = "images/econScatter10.png", width = 800, height = 600) pc6 grid.text("Sources: Transparency International; UN Human Development Report", x = .02, y = .03, just = "left", draw = TRUE) grid.segments(x0 = 0.81, x1 = 0.825, y0 = 0.90, y1 = 0.90, gp = gpar(col = "red"), draw = TRUE) grid.text(paste0("R² = ", as.integer(mR2*100), "%"), x = 0.835, y = 0.90, gp = gpar(col = "gray20"), draw = TRUE, just = "left") dev.off()
Comparing it to the original suggests that we've got most of the important elements, though of course the two graphs are not identical.