Grouping by two variables in ggplot2

Question

I'm trying to create a bar graph using ggplot2 to show the occurrence of different consonants in some language data. For most subjects there is only one timepoint sampled, but for some of the subjects there are two timepoints. The subjects are grouped into two different groups for comparison, and their consonant production is grouped by type - a phonetic variable in my data.

I want to show the consonants produced by each subject in a stacked bar, and then have data from the speakers who were recorded twice show next to each other on the x axis. At the moment, ggplot is aggregating this data. I don't want to facet the data into two different plots.

 exampledata <- tribble(~subject, ~group, ~time, ~consonant, ~type,  ~n,
     "1", "A", 1, "p", F, 10,
     "1", "A", 1, "t", T, 12,
     "1", "A", 1, "k", T, 50,
     "2", "A", 1, "p", T, 0,
     "2", "A", 1, "t", T, 45,
     "2", "A", 1, "k", F, 23,
     "2", "A", 2, "p", F, 2,
     "2", "A", 2, "t", T, 34,
     "2", "A", 2, "k", T, 56,
     "3", "B", 1, "p", F, 12,
     "3", "B", 1, "t", T, 13,
     "3", "B", 1, "k", F, 50,
     "4", "A", 1, "p", T, 10,
     "4", "A", 1, "t", F, 12,
     "4", "A", 1, "k", T, 50,
     "5", "B", 1, "p", T, 0,
     "5", "B", 1, "t", T, 24,
     "5", "B", 1, "k", F, 3,
     "5", "B", 2, "p", F, 23,
     "5", "B", 2, "t", F, 12,
     "5", "B", 2, "k", T, 7,
     "6", "A", 1, "p", F, 52,
     "6", "A", 1, "t", F, 12,
     "6", "A", 1, "k", T, 64
     )

I'm currently using the following code, which generates the attached figure:

 plot1 <- ggplot(data=exampledata, aes(x=subject, y=n, fill=type, colour = group)) +
   geom_bar(stat="identity") +
   scale_fill_manual(values=c("gray97", "gray87")) +
   scale_colour_manual(values = c("royalblue", "navyblue")) +
   theme_bw()
 plot(plot1)
 [![example bar plot][1]][1]

So all I want to do now is somehow create an extra grouping variable that shows time bars side by side.

Roman · Accepted Answer · 2019-07-23 13:05:32Z

2

You can try to work with a facet but removing the faced-like shape

ggplot(exampledata, aes(x= time, y=n, fill=type))  +  
  geom_col() +
  facet_grid(~subject, scales = "free_x", switch = "x") + 
  xlab("subject") + 
  theme_bw() + 
  theme(axis.text.x = element_blank(),
        axis.ticks.x = element_blank(),
        panel.border =element_blank(),
        strip.background = element_blank())

answered Jul 23, 2019 at 13:05

Roman

17.7k3 gold badges39 silver badges52 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Catherine Laing Over a year ago

I was worried that this would look too full-on as I have a lot of subjects in my real dataset, but it looks great! Thank you!

JAQuent · Accepted Answer · 2019-07-23 11:27:27Z

2

Is that what you want?

plot1 <- ggplot(data=exampledata, aes(x=subject, y=n, fill=type, colour = group, group = time)) +
  geom_bar(stat="identity", position = 'dodge') +
  scale_fill_manual(values=c("gray97", "gray87")) +
  scale_colour_manual(values = c("royalblue", "navyblue")) +
  theme_bw()[![enter image description here][1]][1]

answered Jul 23, 2019 at 11:27

JAQuent

1,2541 gold badge13 silver badges28 bronze badges

2 Comments

Roman Over a year ago

The plot looks wrong, right? The FALSE value of subject one is missing.

Catherine Laing Over a year ago

This is almost it but as @Jimbou says, the numbers don't stack correctly

Claudiu Papasteri · Accepted Answer · 2019-07-23 12:30:38Z

2

In @JAQuent's response, although elegante, the stacking is not preserved. I propose this:

exampledata$interaction <- factor(interaction(exampledata$subject, exampledata$time),
                                  levels = sprintf("%d.%d", sort(rep(1:6, 2)), rep(1:2, 6)))

plot1 <- ggplot(data=exampledata, aes(x=interaction, y=n, fill=type, colour = group)) +
  geom_bar(aes(group = subject), stat="identity") +
  scale_fill_manual(values=c("gray97", "gray87")) +
  scale_colour_manual(values = c("royalblue", "navyblue")) +
  theme_bw()
plot(plot1)

edited Jul 23, 2019 at 12:30

answered Jul 23, 2019 at 11:31

Claudiu Papasteri

2,6391 gold badge21 silver badges35 bronze badges

Comments

TimTeaFan · Accepted Answer · 2019-07-23 12:26:00Z

It is a little bit of a hack, but this should do exactly what you want.

First your data needs to contain zeros for all combinations of subject, time, group and type. Like this:

exampledata <- tribble(~subject, ~group, ~time, ~consonant, ~type,  ~n,
                       "1", "A", 1, "p", F, 10,
                       "1", "A", 1, "t", T, 12,
                       "1", "A", 1, "k", T, 50,
                       "1", "A", 2, "p", F, 0,
                       "1", "A", 2, "t", T, 0,
                       "1", "A", 2, "k", T, 0,
                       "2", "A", 1, "p", T, 0,
                       "2", "A", 1, "t", T, 45,
                       "2", "A", 1, "k", F, 23,
                       "2", "A", 2, "p", F, 2,
                       "2", "A", 2, "t", T, 34,
                       "2", "A", 2, "k", T, 56,
                       "3", "B", 1, "p", F, 12,
                       "3", "B", 1, "t", T, 13,
                       "3", "B", 1, "k", F, 50,
                       "3", "B", 2, "p", F, 0,
                       "3", "B", 2, "t", T, 0,
                       "3", "B", 2, "k", F, 0,
                       "4", "A", 1, "p", T, 10,
                       "4", "A", 1, "t", F, 12,
                       "4", "A", 1, "k", T, 50,
                       "4", "A", 2, "p", T, 0,
                       "4", "A", 2, "t", F, 0,
                       "4", "A", 2, "k", T, 0,
                       "5", "B", 1, "p", T, 0,
                       "5", "B", 1, "t", T, 24,
                       "5", "B", 1, "k", F, 3,
                       "5", "B", 2, "p", F, 23,
                       "5", "B", 2, "t", F, 12,
                       "5", "B", 2, "k", T, 7,
                       "6", "A", 1, "p", F, 52,
                       "6", "A", 1, "t", F, 12,
                       "6", "A", 1, "k", T, 64,
                       "6", "A", 2, "p", F, 0,
                       "6", "A", 2, "t", F, 0,
                       "6", "A", 2, "k", T, 0
)

Then you basically draw two bar charts. Normally you would offset each with position_nudge to left and right to have to bars per subject next to each other. However, since the position argument is already taken ("stacked") the hack is to turn subject into a numeric variable, nudge directly with as.numeric(subject) - 0.125 and as.numeric(subject) + 0.125 and then format the x axis accordingly so that noone notices.

ggplot(exampledata,
       aes(fill = type, colour = group)) +
    geom_col(
        data = subset(exampledata, time == 1),
        aes(x = as.numeric(subject) - 0.125, y = n),
        width = 0.25
    ) +
    geom_col(
        data = subset(exampledata, time == 2),
        aes(x = as.numeric(subject) + 0.125, y = n),
        width = 0.25
    ) +
    scale_fill_manual(values=c("gray97", "gray87")) +
    scale_colour_manual(values = c("royalblue", "navyblue")) +
    scale_x_continuous(limits = c(0.5, 6.5),
                       breaks = (c(1:6))) +
    xlab("subjects")

Collectives™ on Stack Overflow

Grouping by two variables in ggplot2

4 Answers 4

1 Comment

2 Comments

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

4 Answers 4

1 Comment

2 Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related