Diagram types with the ggplot2 package

Install required packages
Simulate data
Diagram types
Multi-panel plots
Additional diagram elements
Put different plots on the same page
Save ggplot2 diagrams to file
Further online resources
Detach (automatically) loaded packages (if possible)
Get the article source from GitHub

Install required packages

ggplot2, ggbeeswarm, cowplot, patchwork

wants <- c("ggplot2", "ggbeeswarm", "cowplot", "patchwork")
has   <- wants %in% rownames(installed.packages())
if(any(!has)) install.packages(wants[!has])

Simulate data

Njk    <- 50
P      <- 4
Q      <- 2
IQ     <- rnorm(P*Q*Njk, mean=100, sd=15)
height <- rnorm(P*Q*Njk, mean=175, sd=7)
rating <- factor(sample(LETTERS[1:3], Njk*P*Q, replace=TRUE))
sex    <- factor(rep(c("f", "m"), times=P*Njk))
group  <- factor(rep(c("control", "placebo", "treatA", "treatB"), each=Q*Njk))
sgComb <- interaction(sex, group)
mood   <- round(rnorm(P*Q*Njk, mean=c(85, 80, 110, 90, 130, 100, 93, 121)[sgComb], sd=25))
myDf   <- data.frame(sex, group, sgComb, IQ, height, rating, mood)

Diagram types

Scatterplot

library(ggplot2)
ggplot(myDf, aes(x=height, y=IQ)) +
        geom_point(size=3)

Line plot

group=TRUE is necessary here

(groupM <- aggregate(mood ~ group, data=myDf, FUN=mean))

    group   mood
1 control  86.79
2 placebo 104.29
3  treatA 114.53
4  treatB 105.14

ggplot(groupM, aes(x=group, y=mood, group=TRUE)) +
    geom_point(size=3) +
    geom_line()

(groupM <- aggregate(mood ~ group, data=myDf, FUN=mean))

    group   mood
1 control  86.79
2 placebo 104.29
3  treatA 114.53
4  treatB 105.14

ggplot(groupM, aes(x=group, y=mood, group=TRUE)) +
    geom_point(size=3) +
    geom_step()

Barplot

Absolute frequencies

ggplot(myDf, aes(x=rating)) +
    geom_bar(stat="count") +
        ggtitle("Rating frequencies")

Relative frequencies

ggplot(myDf, aes(x=rating)) +
    geom_bar(stat="count",
             aes(y=(..count..) / sum(..count..))) +
        ggtitle("Relative rating frequencies")

Histogram

Absolute frequencies

ggplot(myDf, aes(x=IQ)) +
        geom_histogram(bins=20) +
        ggtitle("Histogram IQ by group") +
        theme(legend.position="none")

Density with nonparametric kernel density estimator

ggplot(myDf, aes(x=mood)) +
    geom_histogram(aes(y=..density..)) +
    ggtitle("Histogram mood by group") +
    geom_density(color="darkgrey", fill="grey", alpha=0.6)

Boxplot

Add a geom_beeswarm() layer to show raw data.

library(ggbeeswarm)
ggplot(myDf, aes(x=sex, y=height)) +
    geom_boxplot(outlier.shape=NULL) +
    geom_beeswarm(alpha=0.5) +
    ggtitle("Height by sex") +
    theme(legend.position="none")

Q-Q plot

Against normal distribution

ggplot(myDf, aes(sample=height)) +
    geom_qq() +
    geom_qq_line(color="blue")

Against pre-specified distribution (here: t)

parL <- list(df=10, ncp=0)
ggplot(myDf, aes(sample=height)) +
    geom_qq(distribution=qt, dparams=parL) +
    geom_qq_line(color="blue", distribution=qt, dparams=parL)

Compare quantiles of two variables

probs <- ppoints(50)
qDat  <- data.frame(qmood=quantile(myDf$mood, probs),
                    qheight=quantile(myDf$height, probs))

ggplot(qDat, aes(x=qmood, y=qheight)) +
    geom_point() +
    xlab("Quantiles mood") +
    ylab("Quantiles height")

Multi-panel plots

Split the diagram into panels by column using facet_grid().

ggplot(myDf, aes(x=rating, y=IQ)) +
        geom_point(size=3) +
        facet_grid(. ~ group) +
        ggtitle("IQ ~ height split by sex and group") +
        guides(shape="none")

Split the diagram into matrix of panels using facet_wrap(). Option scales="free_y" produces separate y-axis for each panel.

ggplot(myDf, aes(x=rating, y=IQ)) +
        geom_point(size=3) +
        facet_wrap(. ~ group, scales="free_y") +
        ggtitle("IQ ~ height split by sex and group") +
        guides(shape="none")

Additional diagram elements

Add error bars

groupN     <- as.data.frame(xtabs(~ sex + group, data=myDf))
groupM     <- aggregate(mood ~ sex + group, data=myDf, FUN=mean)
groupSD    <- aggregate(mood ~ sex + group, data=myDf, FUN=sd)
groupMSD   <- merge(groupM, groupSD, by=c("sex", "group"), suffixes=c(".M", ".SD"))
groupMSDN  <- merge(groupMSD, groupN, by=c("sex", "group"))
(groupMSDN <- transform(groupMSDN,
                        SEMlo=mood.M - mood.SD/sqrt(Freq),
                        SEMup=mood.M + mood.SD/sqrt(Freq)))

  sex   group mood.M  mood.SD Freq     SEMlo     SEMup
1   f control  90.16 26.10916   50  86.46761  93.85239
2   f placebo 115.32 22.71064   50 112.10823 118.53177
3   f  treatA 125.52 27.40791   50 121.64394 129.39606
4   f  treatB  91.36 25.60035   50  87.73956  94.98044
5   m control  83.42 25.34826   50  79.83521  87.00479
6   m placebo  93.26 23.49452   50  89.93737  96.58263
7   m  treatA 103.54 23.58416   50 100.20470 106.87530
8   m  treatB 118.92 22.67772   50 115.71289 122.12711

ggplot(groupMSDN, aes(x=group, y=mood.M, ymin=SEMlo, ymax=SEMup,
                              color=sex, shape=sex, group=sex)) +
    geom_point() +
    geom_line() +
    geom_linerange()

geom_hline() for a horizontal line
geom_vline() for a vertical line
geom_smooth() for a regression line including SEs
geom_text() for text from a variable
ggtitle() for a diagram title
annotate() for text given directly as an argument

ggplot(myDf, aes(x=height, y=IQ)) +
        geom_hline(aes(yintercept=100), linetype=2) +
        geom_vline(aes(xintercept=180), linetype=2) +
        geom_point(size=3) +
        geom_smooth(method=lm, se=TRUE, size=1.2, fullrange=TRUE) +
        facet_grid(sex ~ group) +
        ggtitle("IQ ~ height split by sex and group") +
        theme(legend.position="none") +
        geom_text(aes(x=190, y=70, label=sgComb)) +
        annotate("text", x=165, y=130, label="annotation")

Put different plots on the same page

Using package cowplot

library(cowplot)
p1 <- ggplot(myDf, aes(x=height, y=IQ)) +
    geom_point()

p2 <- ggplot(myDf, aes(x=rating)) +
    geom_bar(stat="count", position=position_stack()) +
    ggtitle("Rating frequencies by sex and group")

plot_grid(p1, p2, ncol=1, align="v")

Using package patchwork

library(patchwork)
p1 + p2 + plot_layout(ncol=2)

p1 + p2 + plot_layout(ncol=1)

Save `ggplot2` diagrams to file

Graphics file format is determined from file ending.

p <- ggplot(myDf, aes(x=height, y=IQ)) +
        geom_point(size=3)

ggsave(p, file="diag_ggplot.pdf", width=9, height=6, units="in")
ggsave(p, file="diag_ggplot.png", width=200, height=150, units="mm", dpi=300)

Further online resources

See Cookbook for R: ggplot2 diagrams for many detailed examples of ggplot2 diagrams.
Package esquisse provides an interactive Shiny app to create and customize ggplot2 diagrams.

Detach (automatically) loaded packages (if possible)

try(detach(package:cowplot))
try(detach(package:patchwork))
try(detach(package:ggbeeswarm))
try(detach(package:ggplot2))

Get the article source from GitHub

R markdown - markdown - R code - all posts