ggBubbles

In ggBubbles: Mini Bubble Plots for Comparison of Discrete Data with 'ggplot2'

Introduction

Welcome to this vignette!

Loading the package

require(knitr)

You can load the package with this command

require(ggplot2)
require(ggBubbles)

In 15 seconds

The package introduces position_surround() for r CRANpkg("ggplot2").

Parameter is offset which controls the offsets for position corrections (default is 0.1).

position_surround() can be used in many r CRANpkg("ggplot2") functions like geom_point or geom_text:

suppressPackageStartupMessages({
    require(dplyr)
    require(tibble)
})
data(MusicianInterestsSmall)
ggplot(data = MusicianInterestsSmall,
       aes(x = Instrument,
           y = Genre,
           col = Level)) +
       geom_point(position = position_surround(), size = 4) +
       scale_colour_manual(values = c("#00e5ff",
                                      "#4694ff",
                                      "#465aff",
                                      "#2c00c9")) + 
       theme_bw(base_size = 18)

Bubbleplot vs Minibubble plot

Here we demonstrate the advantage of MiniBubble plots compared to traditional Bubbleplot in certain usecases with discrete data.

Please not that in this vignette we will use r CRANpkg("dplyr") and r CRANpkg("tibble") from r CRANpkg("tidyverse").

require(dplyr)
require(tibble)

Example data

First, we load a small example data

data(MusicianInterestsSmall)

which contains data from musicians about their experience in differente music genres they have with their music instruments.

head(MusicianInterestsSmall)

kable(head(MusicianInterestsSmall),
      caption = "First rows of MusicianInterestsSmall")

Traditional Bubble plot

The traditional bubble plot is able to portrait the amount of guitarrists or pianists able to play jazz or classical music by size and display the average experience level by colour coding.

ggplot(data = MusicianInterestsSmall %>% 
               group_by(Instrument, Genre) %>% 
               summarize(Count = n(), AvgLevel = mean(as.integer(Level))),
        aes(x = Instrument, y = Genre, size = Count, col = AvgLevel)) +
        geom_point() + theme_bw(base_size = 18) +
        scale_colour_gradientn(
            colours  = rev(topo.colors(2)),
            na.value = "transparent",
            breaks   = as.integer(MusicianInterestsSmall$Level) %>% 
                             unique %>% sort,
            labels   = levels(MusicianInterestsSmall$Level),
            limits   = c(as.integer(MusicianInterestsSmall$Level) %>% min,
                         as.integer(MusicianInterestsSmall$Level) %>% max)) +
        scale_size_continuous(range = c(3, 11)) 

From a data visualisation point of view, it is debateable how good point sizes are to display counts. However, in general we can agree that averages often hide a lot of useful information.

MiniBubble Plot

The MiniBubble plot allows to show each musician and their corresponding skill level individually:

ggplot(data = MusicianInterestsSmall,
       aes(x = Instrument,
           y = Genre,
           col = Level)) +
       geom_point(position = position_surround(), size = 4) +
       scale_colour_manual(values = c("#00e5ff",
                                      "#4694ff",
                                      "#465aff",
                                      "#2c00c9")) + 
       theme_bw(base_size = 18)

This is done by the position_surround() function passed to the position argument of geom_point. Note, that only exact overlaps will be dodged. The points will surround the center in layers which will be filled clockwise.

Since each individual data point is shown seperately, you can also use shape and fill to show further features, as long the plot will not be overloaded with information.

Also, you can use geom_text(position = position_surround()) to overlay the points with text, or make the text appear in shiny when hovering.

MiniBubbleplot allows to show more features in a bubble plot.

Offset parameter

The offset of the dodged points can be handed as parameters to position_surround().

ggplot(data = MusicianInterestsSmall,
       aes(x = Instrument,
           y = Genre,
           col = Level)) +
       geom_point(position = position_surround(offset = .2), size = 4) +
       scale_colour_manual(values = c("#00e5ff",
                                      "#4694ff",
                                      "#465aff",
                                      "#2c00c9")) + 
       theme_bw(base_size = 18)

Another example

We load a bigger test data set:

data(MusicianInterests)

This dataset also contains information about the musicians themselves from the multiple - choice survey.

head(MusicianInterests)

kable(head(MusicianInterests),
      caption = "First rows of MusicianInterests")

The basis of the plot is simply:

p <- ggplot(data        = MusicianInterests,
            aes(x       = Genre,
                y       = Instrument,
                col     = Level)) +
    geom_point(size     = 1.8, 
               position = position_surround(offset = .2))

Here we add some graphical parameters to make it pretty:

p <- p + theme_bw(base_size = 17) + 
    theme(axis.text.x = element_text(angle = 45, hjust = 1)) +
    scale_colour_gradientn(
            colours  = rev(topo.colors(2)),
            na.value = "transparent",
            breaks   = 1:6,
            labels   = c("Interested",
                         "Beginner",
                         "Intermediate",
                         "Experienced",
                         "Very experienced",
                         "Pro")) +
    xlab("") + ylab("")
p

Algorithmic details

The position_surround() algorithm determines how many points are overlaying and then displays the points in clockwise ration around the center in quadratic layers. Please see the graphical illustration here:

n <- 25
X <- data.frame(x = rep(1, sum(1:n)),
           y = rep(1, sum(1:n)),
           group = unlist(lapply(1:n, function(x) { rep(x, x) })),
           label =  unlist(lapply(1:n, function(x) { 1:x }))
)

ggplot(data = X, 
       aes(x = x, y = y, label = label)) +
       facet_wrap(~group) + 
       geom_text(position = position_surround(offset=.4)) + 
       theme_minimal() + 
       theme( 
           strip.background = element_blank(),
          strip.text.x = element_blank()
       ) + 
       xlim(c(0,2)) + ylim(c(0,2))

Support or Feedback

For feedback or suggestions please contact the maintainer: Thomas Schwarzl [e-mail] or [e-mail].

ggBubbles documentation built on Sept. 4, 2019, 9:02 a.m.