x2-Graphics with R

Applied Statistics – A Practical Course

Thomas Petzoldt

2024-08-23

Preface

The examples were tested with R 4.3.1

Data sets of this presentation:

Some data sets are built-in data in R
Other data can be downloaded from https://github.com/tpetzoldt/datasets

The `radiation.csv’ data set contains derived data from the German Weather Service (http://www.dwd.de), station Dresden-Klotzsche. Missing data were interpolated.

If a data set is missing, please let me know.

The easy way
Customizing graphics
Multiple grapics on one page
Saving and exporting graphics
Lattice
grid and gridbase
ggplot2

Base graphics

The Easy Way

plot(iris)

R contains many graphics functions with convenient defaults.
iris is a built-in data set in R (see next slide)
plot is a so-called generic function that automatically decides how to plot.

The iris data set

The famous (Fisher’s or Anderson’s) iris data set contains measurements (in centimeter) of the variables sepal length, sepal width, petal length and petal width of 50 flowers from each of 3 species of iris, Iris setosa, I. versicolor, and I. virginica.

see ?iris in R’s online help.
or: https://en.wikipedia.org/wiki/Iris_flower_data_set

Plotting colums of a data frame

plot(iris$Sepal.Length, iris$Petal.Length)

plot(iris$Sepal.Length, iris$Petal.Length,
     col=iris$Species)

A column of a data.frame is accessed with $.

The use of `with()` saves dollars

plot(iris$Sepal.Length, iris$Petal.Length, 
     col=iris$Species)

with(iris, plot(Sepal.Length, Petal.Length, 
                col=Species))

Colors and plotting symbols in R

R allows to change style and color of plotting symbols:

col: color, can be one of 8 default colors or a user-defined color
pch: plotting character, can be one of 25 symbols or a quoted letter
cex: character extension: size of a plotting character

plot(1:25, col=1:25, pch=1:15, cex=2)

Special plotting symbols

symbols 21..25 have an optional background color
lwd: border width of the symbol

plot(21:25, col="darkred", pch=21:25, cex=2, bg="green", lwd=2)

R as function plotter

x <- seq(0, 20, length.out=100)
y1 <- sin(x)
y2 <- cos(x)
plot(x, y1, type="l", col="red")
lines(x, y2, col="blue")

type: “p”: points, “l”: lines, “b”: both, points and lines, “c”: empty points joined by lines, “o”: overplotted points and lines, “s” and “S”: stair steps, “h” histogram-like vertical lines, “n”: no points or lines.

Line styles

x <- seq(0, 20, length.out=100)
plot(x, sin(x), type="l", col="red", lwd=3, lty="dotted")
lines(x, cos(x), col="blue", lwd=2, lty="dashed")

lty: line type (“blank”, “solid”, “dashed”, “dotted”, “dotdash”, “longdash”, “twodash”) or a number from 1…7, or a string with up to 8 numbers for drawing and skipping (e.g. “4224”).
lwd: line width (a number, defaults to 1)

Coordinate axes and annotations

plot(iris$Sepal.Length, iris$Petal.Length, xlim=c(0, 8), ylim=c(2,8),
     col=iris$Species, pch=16,
     xlab="Sepal Length (cm)", ylab="Petal Length (cm)", main="Iris Data",
     las = 1)

col=iris$Species: works because Species is a factor
las=1: numbers on y-axis upright (try: 0, 1, 2 or 3)
log: may be used to transform axes (e.g. log=“x”, log=“y”, log=“xy”)

Adding a legend

mycolors <- c("blue", "red", "cyan")
plot(iris$Sepal.Length, iris$Petal.Length, xlim=c(0, 8), ylim=c(2,8),
     col=mycolors[iris$Species], pch = 16,
     xlab="Sepal Length (cm)", ylab="Petal Length (cm)", main="Iris Data",
     las = 1)
legend("topleft", legend=c("Iris setosa", "Iris versicolor", "Iris virginica"),
  col=mycolors, pch=16)

see ?legend for more options (e.g. line styles, position of the legend)

Global parameters, font size, margins, …

Many figure options can be specified globally with par()
par(lwd=2) all lines have double width
par(mfrow=c(2,2)) subdivides the graphics area in 2 x 2 fields
par(las=1) numbers at y axis upright
par(mar=c(5, 5, 0.5, 0.5)) changes figure margins (bottom, left, top, right)
par(cex=2) increase font size
$\rightarrow$ sometimes it is better to leave font size as is and change size of the figure instead

Read the ?par help page!

Example

#
plot(iris$Sepal.Length, iris$Petal.Length, 
     col=iris$Species)
#

opar <- par(cex=2, mar=c(4,4,1,1), las=1)
plot(iris$Sepal.Length, iris$Petal.Length, 
     col=iris$Species)
par(opar)

change font size, margins and axis label orientation
opar stores previuos parameter and allows resetting

Saving and exporting figures

Easiest way ist to use the RStudio’s Export –> Save as Image (or copy to clipboard)
Important: Select correct image format and image size!

Format	Type	Usage	Notes
PNG	bitmap	general purpose	fixed size, use at least 300 pixels per inch
JPEG	bitmap	photographs	not good for R images
TIFF	bitmap	PNG is easier	outdated, required by some journals
BMP	bitmap	not recommended	outdated, needs huge memory
Metafile	vector	Windows standard format	easy to use, quality varies
SVG	vector	can be edited	allows editing graphics with Inkscape
EPS	vector	PDF is easier	required by some journals
PDF	vector	best quality	perfect for LaTex, RMarkdown and Quarto, MS Office requires conversion

Writing figures directly to PDF

pdf("myfile.pdf", width=8, height=6)
par(las=1)
plot(iris$Sepal.Length, iris$Petal.Length, col=iris$Species)
dev.off()

width and height in inch (1 inch = 2.54cm)
professional quality, size can be changed without quality loss
conversion to PNG can be done later with free programs

$\rightarrow$ Inkscape, SumatraPDF, ImageMagick

Writing figures directly to PNG

png("myfile.png", width=1600, height=1200, res=300)  # good for Word
#png("myfile.png", width=800, height=600, res=150)   # good for Powerpoint
par(las=1)
par(mar=c(5, 5, 1, 1))
plot(iris$Sepal.Length, iris$Petal.Length)
dev.off()

width and height given in pixels
Hint: play with res to change nominal resolution and font size
use at least 300 dpi (dots per inch, i.e. number of pixels = 300/2.54 * width in cm)
professionals use 600 or even 1200 pixels per inch, but then .docx and .pptx files will dramatically increase
1600 x 1200px is good for 13.3 x 10 cm size in the printed document

Example: Solar Radiation Data in Dresden

radiation <- read.csv("../data/radiation.csv")
radiation$Date <- as.Date(radiation$date)

plot(radiation$Date, radiation$rad)

Note: The data set contains derived data from the German Weather Service (http://www.dwd.de), station Dresden. Missing data were interpolated.

Some notes About Date and Time

Date and time classes in R

Most important classes
- as.Date (dates only)
- as.POSIXct (date and time)
format and strptime
extract day, month, year, Julian day
time series objects tseriesand zoo

format and strptime

format(x, format = "", tz = "", usetz = FALSE, ...)

%Y	year with century
%m	month as decimal number
%d	day of the month
%H	hours as decimal number (00-23)
%M	minute as decimal number (00-59)
%S	second as decimal number (00-59)
%j	day of year (001-366)
%u	weekday, Monday is 1

as.Date("11.03.2015", format="%d.%m.%Y")

[1] "2015-03-11"

Date conversion for the solar radiation data set

radiation$year <- format(radiation$Date, "%Y")
radiation$month <- format(radiation$Date, "%m")
radiation$doy <- format(radiation$Date, "%j")
radiation$weekday <- format(radiation$Date, "%u")

head(radiation)

        date rad interpolated       Date year month doy weekday
1 1981-01-01 197            0 1981-01-01 1981    01 001       4
2 1981-01-02  89            0 1981-01-02 1981    01 002       5
3 1981-01-03  49            0 1981-01-03 1981    01 003       6
4 1981-01-04 111            0 1981-01-04 1981    01 004       7
5 1981-01-05 161            0 1981-01-05 1981    01 005       1
6 1981-01-06  55            0 1981-01-06 1981    01 006       2

The lubridate package has date and time functions that are easier to use.

Summarize data with `aggregate`

Syntax

aggregate(x, by, FUN, ..., simplify = TRUE)

Example

yearmax <- aggregate(
  list(rad = radiation$rad),
  list(year = radiation$year),
  max)

monmean <- aggregate(
  list(radiation = radiation$rad),
  list(year = radiation$year, month = radiation$month),
  mean)

aggregate is essentially a wrapper to apply

Plot aggregated radiation data

par(mfrow=c(1,2), las=1)
boxplot(rad ~ year, data = radiation)
boxplot(rad ~ month, data = radiation)

Most functions that support a formula argument (containing ~) allow to specify the data frame with a data argument.

More about Plotting

Different plotting packages with different philosophies

base graphics
package lattice
package ggplot2
Manipulation of plots
- set size and fonts; save plots to disk
- use pdf, svg or png – not jpg - except for photographs
Related software
- edit/convert svg (and pdf) with Inkscape
- convert images with ImageMagick

Base Graphics

x <- rnorm(100)
par(mfrow=c(2,2))
plot(x)
hist(x)
qqnorm(x)
boxplot(x)

grid and gridBase

complete freedom to organise plotting area
interface relatively raw
basis of other plotting packages

Lattice Graphics

Implements “trellis graphics” (i.e. gridded graphics) in R
Sarkar, D. (2008). Lattice: multivariate data visualization with R. Springer Science & Business Media.

require(lattice)
data(iris)
xyplot(Sepal.Length ~ Sepal.Width|Species, data=iris, layout=c(3,1))

ggplot2

Implements the “Grammar of Graphics”
- Leland Wilkinson (2005) The Grammar of Graphics. 2nd edn. Springer
- Hadley Wickham (2009, 2016) ggplot2: Elegant Graphics for Data Analysis. Springer.
very popular, part of the tidyverse family of packages
https://ggplot2.tidyverse.org/

`ggplot`-Example

library(ggplot2)
data(iris)
ggplot(iris, aes(Sepal.Length, Sepal.Width, color = factor(Species))) + 
  geom_point() + 
  stat_smooth(method = "lm")

Pipelines and faceting in ggplot2

library("dplyr")
library("lubridate")
library("ggplot2")
read.csv("../data/radiation.csv") |>
  mutate(year=year(date), doy=yday(date)) |>
  ggplot(aes(doy, rad)) + geom_line() + facet_wrap(. ~ year)

Save Images to Disk

Simple: use of RStudio’s graphics export
Better:

pdf(file="figure.pdf",width=8,height=6) # open
plot(rnorm(100))                        # plotting
dev.off()                               # close

other graphics formats: png(), svg(), …
pdf allows several plots in one file (until dev.off)
special formats possible, e.g. width=100, height=10 (useful for hclust)
in case of error: repeated dev.off() to close all files.

Vector vs. Bitmap Graphics

Bitmap formats

jpg, png, tiff
fixed resolution, cannot be magnified without loss
retouching possible, but not editing
well suited for pictures or plots with huge number of data (color maps)
cannot be converted to vector without complications and quality loss

Vector formats

svg, pdf, [wmf, emf]
can be up- and downscaled and edited
well suited drawings and diagrams (except if huge amount of data)
can always be converted to bitmap

x2-Graphics with R

Preface

Table of Contents

Base graphics

The Easy Way

The iris data set

Plotting colums of a data frame

The use of `with()` saves dollars

Colors and plotting symbols in R

Special plotting symbols

R as function plotter

Line styles

Coordinate axes and annotations

Adding a legend

Global parameters, font size, margins, …

Example

Saving and exporting figures

Writing figures directly to PDF

Writing figures directly to PNG

Example: Solar Radiation Data in Dresden

Some notes About Date and Time

Date and time classes in R

format and strptime

Date conversion for the solar radiation data set

Summarize data with `aggregate`

Plot aggregated radiation data

More about Plotting

Different plotting packages with different philosophies

Base Graphics

grid and gridBase

Lattice Graphics

ggplot2

`ggplot`-Example

Pipelines and faceting in ggplot2

Save Images to Disk

Vector vs. Bitmap Graphics

Further Reading

x2-Graphics with R

Preface

Table of Contents

Base graphics

The Easy Way

The iris data set

Plotting colums of a data frame

The use of with() saves dollars

Colors and plotting symbols in R

Special plotting symbols

R as function plotter

Line styles

Coordinate axes and annotations

Adding a legend

Global parameters, font size, margins, …

Example

Saving and exporting figures

Writing figures directly to PDF

Writing figures directly to PNG

Example: Solar Radiation Data in Dresden

Some notes About Date and Time

Date and time classes in R

format and strptime

Date conversion for the solar radiation data set

Summarize data with aggregate

Plot aggregated radiation data

More about Plotting

Different plotting packages with different philosophies

Base Graphics

grid and gridBase

Lattice Graphics

ggplot2

ggplot-Example

Pipelines and faceting in ggplot2

Save Images to Disk

Vector vs. Bitmap Graphics

Further Reading

The use of `with()` saves dollars

Summarize data with `aggregate`

`ggplot`-Example