Side-by-side examples
The ggmosaic paper (Jeppson & Hofmann, 2023, The R
Journal 14(4)) uses the fly dataset from ggmosaic and
the built-in Titanic dataset. Since ggmosaic is no longer
on CRAN, we recreate the fly-based examples using
Titanic and HairEyeColor — both built-in R
datasets that need no extra packages.
One dimensional — spine plot and bar chart
The paper starts with single-variable plots showing different divider types (Figure 1 in Jeppson & Hofmann, 2023):
# ggmosaic — spine plot (one variable, default divider)
ggplot(data = titanic) +
geom_mosaic(aes(x = product(Class), fill = Class, weight = Freq))# marimekko — one variable
ggplot(titanic) +
geom_marimekko(aes(fill = Class, weight = Freq), formula = ~Class) +
theme_marimekko() +
labs(title = "One variable: f(Class)")
# ggmosaic — bar chart (one variable, hbar divider)
ggplot(data = titanic) +
geom_mosaic(
aes(x = product(Class), fill = Class, weight = Freq),
divider = "hbar"
)# ggplot — barchart
# todo: needs percentage y scale
ggplot(titanic) +
geom_bar(aes(Class, weight = Freq, fill = Class)) +
theme_marimekko() +
labs(title = "One variable: f(Class)")
The main difference: no product()
wrapper — use formula = ~ Var instead. marimekko
does not have a divider argument; it always produces
vertical columns with horizontal stacking.
Two dimensional — mosaic plot and stacked bar chart
The paper shows two-dimensional mosaic plots where variable order in
product() controls the partitioning hierarchy (Figure
2):
# ggmosaic — two variables
ggplot(data = titanic) +
geom_mosaic(aes(
x = product(Class),
fill = Survived,
weight = Freq
))# marimekko — two variables via formula
ggplot(titanic) +
geom_marimekko(aes(fill = Survived, weight = Freq),
formula = ~ Class | Survived
) +
theme_marimekko() +
labs(title = "Two variables: f(Survived | Class) f(Class)")
Three dimensional
The paper shows two-dimensional mosaic plots where variable order in
product() controls the partitioning hierarchy (Figure
3):
# ggmosaic — double-decker variables
ggplot(data = titanic) +
geom_mosaic(aes(
x = product(Survived, Class, Sex),
fill = Survived,
weight = Freq,
), divider = ddecker())# marimekko — three variables via formula
ggplot(titanic) +
geom_marimekko(aes(fill = Survived, weight = Freq),
formula = ~ Sex + Class | Survived
) +
theme_marimekko() +
labs(title = "Three variables: f(Survived | Class, Sex)")
Three variables — nested mosaic
The paper demonstrates three-variable mosaics where a third variable adds another level of partitioning (Figure 4):
# ggmosaic — three variables via product()
ggplot(data = titanic) +
geom_mosaic(aes(
x = product(Sex, Survived, Class),
fill = Survived,
weight = Freq
))# marimekko — three variables via formula
ggplot(titanic) +
geom_marimekko(aes(fill = Survived, weight = Freq),
formula = ~ Class | Survived | Sex
) +
theme_marimekko() +
labs(title = "Three variables: Class / Sex / Survived")
Conditioning with conds
The paper shows how conds creates conditional
distributions (Section 3.2):
# ggmosaic — conditioning
ggplot(data = titanic) +
geom_mosaic(aes(
x = product(Class),
fill = Survived,
weight = Freq,
conds = product(Sex)
))marimekko does not have a conds aesthetic. Use
facet_wrap() instead:
# marimekko — conditioning via faceting
ggplot(titanic) +
geom_marimekko(aes(fill = Survived, weight = Freq),
formula = ~ Class | Survived
) +
facet_wrap(~Sex) +
theme_marimekko() +
labs(title = "f(Survived | Class) conditioned on Sex")
Faceting
The paper demonstrates facet_grid() as an alternative to
conditioning (Section 3.3):
# ggmosaic — faceting
ggplot(data = titanic) +
geom_mosaic(aes(
x = product(Class), fill = Survived, weight = Freq
)) +
facet_grid(~Sex)# marimekko — faceting works the same way
ggplot(titanic) +
geom_marimekko(aes(fill = Survived, weight = Freq),
formula = ~ Class | Survived
) +
facet_wrap(~Sex) +
theme_marimekko() +
labs(title = "Mosaic faceted by Sex")
Divider types and standardize
The paper describes four fundamental partition types:
hspine, vspine, hbar, and
vbar (Section 4, Figure 5). These are combined to produce
spine plots, stacked bar charts, mosaic plots, and double decker
plots.
marimekko simplifies this — geom_marimekko() always
starts with horizontal splits and alternates direction with each
| in the formula. Use coord_flip() if you need
vertical-first orientation.
For equal-width columns (spine-plot style),
standardize = TRUE is available in
geom_marimekko_text() and
fortify_marimekko().
# ggmosaic — spine plot via divider
ggplot(titanic) +
geom_mosaic(aes(x = product(Class), fill = Survived, weight = Freq),
divider = c("vspine", "hspine")
)# marimekko — proportional-width columns (default)
ggplot(titanic) +
geom_marimekko(aes(fill = Survived, weight = Freq),
formula = ~ Class | Survived
) +
theme_marimekko() +
labs(title = "Mosaic plot (proportional-width columns)")
Offset / spacing
The paper shows how offset controls gaps between tiles
(Section 5):
# ggmosaic — single offset parameter
ggplot(data = titanic) +
geom_mosaic(aes(
x = product(Class), fill = Survived, weight = Freq
), offset = 0.02)marimekko provides independent gap_x and
gap_y parameters:
# marimekko — no gaps
ggplot(titanic) +
geom_marimekko(aes(fill = Survived, weight = Freq),
formula = ~ Class | Survived, gap_x = 0.02, gap_y = 0.02
) +
theme_marimekko() +
labs(title = "No gaps")
Text labels
The paper shows geom_mosaic_text() for adding labels
(Section 7):
# ggmosaic — automatic text labels
ggplot(titanic) +
geom_mosaic(aes(x = product(Class), fill = Survived, weight = Freq)) +
geom_mosaic_text(aes(x = product(Class), fill = Survived, weight = Freq))# marimekko — automatic tile positions, only label needed
ggplot(titanic) +
geom_marimekko(aes(fill = Survived, weight = Freq),
formula = ~ Class | Survived
) +
geom_marimekko_text(aes(
label = after_stat(paste(Class, Survived, sep = "\n"))
)) +
theme_marimekko()
geom_marimekko_text() reads tile positions from the
preceding geom_marimekko() layer — only the
label aesthetic is needed. Use after_stat()
for computed variables like weight (count),
.proportion, or the original variable columns.
Axis labels and percentages
The paper uses scale_x_productlist() for axis
labels:
# ggmosaic
ggplot(titanic) +
geom_mosaic(aes(x = product(Class), fill = Survived, weight = Freq)) +
scale_x_productlist()# marimekko — with optional marginal percentages
ggplot(titanic) +
geom_marimekko(aes(fill = Survived, weight = Freq),
formula = ~ Class | Survived,
show_percentages = TRUE
) +
theme_marimekko()
HairEyeColor — a larger example
The paper uses multi-variable examples to show how mosaic plots
reveal associations. Here we use the built-in HairEyeColor
dataset:
ggplot(hair) +
geom_marimekko(aes(fill = Eye, weight = Freq),
formula = ~ Hair | Eye,
show_percentages = TRUE
) +
theme_marimekko() +
labs(title = "Hair color vs Eye color")
Three-variable version with Sex:
ggplot(hair) +
geom_marimekko(aes(fill = Eye, weight = Freq),
formula = ~ Hair | Sex | Eye
) +
theme_marimekko() +
labs(title = "Hair / Sex / Eye")