Introduction

Mosaicplot is a useful way to visualize multivariate categorical data. In R, there are several functions from different packages to draw mosaicplot. In this analysis, I will compare the geom_mosaic function from ggmosaic package and the mosaic function from vcd function. I will use the built-in dataset called Titanic to draw the plots.

Data description

The Titanic dataset summarizes the information on the fate of passengers on the fatal maiden voyage of the ocean liner Titanic. In this dataset, there are four variables, which are Class, Sex, Age and Survived. For Class variable, it records which class a passenger belongs to and has four levels, which are first class, second class, third class, and crew members. For Sex variable, it has two categories, which are male and female. For Age variable, it has two categories, which are child and adult. For Survived categories, it records whether the passager survived or not.

The Titanic dataset in R is a 4-dimensional array. it records the number of people in each combination of the four variables. For example, the number of children who did not survive, are male, and belong to first class is 0.

The Titanic dataset are shown below.

Titanic
## , , Age = Child, Survived = No
## 
##       Sex
## Class  Male Female
##   1st     0      0
##   2nd     0      0
##   3rd    35     17
##   Crew    0      0
## 
## , , Age = Adult, Survived = No
## 
##       Sex
## Class  Male Female
##   1st   118      4
##   2nd   154     13
##   3rd   387     89
##   Crew  670      3
## 
## , , Age = Child, Survived = Yes
## 
##       Sex
## Class  Male Female
##   1st     5      1
##   2nd    11     13
##   3rd    13     14
##   Crew    0      0
## 
## , , Age = Adult, Survived = Yes
## 
##       Sex
## Class  Male Female
##   1st    57    140
##   2nd    14     80
##   3rd    75     76
##   Crew  192     20

Examples using geom_mosaic function from ggmosaic package

At first, I will use geom_mosaic function to draw the mosaicplot. Before drawing the mosaicplot, we need to convert the dataset into data frame since ggplot only accepts data frame. If we directly use the dataset in an array form, R will give us an error, which is “data must be a data frame, or other object coercible by fortify(), not an S3 object with class table”.

library(ggplot2)
library(ggmosaic)
#convert array into data frame
Titanic.df <- data.frame(Titanic)
ggplot(data = Titanic.df) +
  geom_mosaic(aes(weight=Freq, x=product(Class, Sex), fill=Survived), divider = mosaic("v"))

The mosaicplot above shows the Survived variable depends on class and sex. In the geom_mosaic function, I define a aesthetic that uses the frequency for the weight of each combination and specifies class and sex as independent variables. I also specify different color to represent the dependent variable by using fill argument. divider argument specifies the direction of the first split is vertical.

It is worth to notice that, we need to set aesthetic in the geom_mosaic() instead of ggplot(). If we set aesthetic in ggplot(), we will get an error, which is “mapping must be created by aes()”.

ggplot(data = Titanic.df) +
  geom_mosaic(aes(weight=Freq, x=product(Class, Sex, Age), fill=Survived), divider = mosaic("v"))

The mosaicplot above shows the Survived variable depends on class, sex, and age.

Examples using mosaic function from vcd package

Next, I will use mosaic function to draw the mosaicplot. Unlike ggplot, the mosaic function can directly take the data in an array form, which is convenient.

vcd::mosaic(Survived~Sex+Class, data = Titanic, direction = c("v","v","h"), rot_labels = c(0, 0, 45, 0))

The mosaicplot above shows the Survived variable depends on class and sex. In the mosaic function, I define a formula to specify the variables used to create a contingency table from data argument. For the formula, I specify a dependent variable before the tilde sign, and the variables after the tilde sign will be independent variables. Then I specify the direction of each split, “v” for vertical split and “h” for horizontal split. We can use rot_labels to solve overlap issue beween labels.

vcd::mosaic(Survived~Sex+Class+Age, data = Titanic, direction = c("v","v","h","h"), rot_labels = c(0, 0, 45, 0))

The mosaicplot above shows the Survived variable depends on class, sex, and age.

Comparison

For geom_mosaic():

  1. we can only specify the direction of the first cut.

  2. we should to convert data into data frame, when we use ggplot.

  3. we can only set the aesthetic in geom_mosaic() instead of ggplot(). It is kind of violating the convention of ggplot2 package since for other function we can set aesthetic in either geom funciton or ggplot().

  4. it will only use bottom and left axles to make each split, which will become more clear for us to find patterns when there are more than 2 independent variables, just like the second plot drawn by geom_mosaic function.

  5. geom_mosaic cannot deal with a number of variables. The solution for this issue is using the product funciton when we set the value for x in the aesthetic. [refer: https://cran.r-project.org/web/packages/ggmosaic/vignettes/ggmosaic.html]

For mosaic():

  1. we can use direction argument to specify the direction of each cut.

  2. The data argument can accept data frame or array.

  3. There is rot_labels argument in the funciton, thus we can easily rotate labels by certain angle when there is overlap between labels, but for geom_mosaic, we need to use other ggplot2 function to rotate labels manually.

  4. mosaic function will use all 4 axles to make each split. when there are only one or two independent variables, we will put independent variables in the top and bottom axles and dependent varialbe in the horzontal axis. It will become much easier to find the pattern or relationship between variables; however, if there are more than two independent variables, the mosaicplot drawn by mosaic function will become confused and hard to find pattern, just like the second plot drawn by mosaic function.