Data Visualization with ggplot2

Data Visualization in R

R is nowadays equipped with two independent (incompatible, yet coexisting) systems for graphics generation:

• Traditional System – base graphics
  • Built-in functionality for creating charts and graphs in R.
  • Easier to learn and use.
  • Discussed before.
• Newer System – grid
  • ggplot2 – an open-source data visualization package.
  • Very flexible and gaining popularity rapidly nowadays.
  • Very brief introduction to ggplot in this lecture.
  • R Graph Gallery https://r-graph-gallery.com/
  • ggplot Cheatsheet https://rstudio.github.io/cheatsheets/html/data-visualization.html

Five Basic Graphs

Visualization using ggplot2 will be introduced and compared with Base R Graphics.

• scatterplot
• linegraph
• histogram
• boxplot
• barplot

Recall: Base R Graphics

Scatterplot Example

Example: Draw scatterplot of wind vs pressure of the data set storms (in the tidyverse package).

# Plot scatter plot with x = pressure, y = wind
plot(storms$pressure, storms$wind, col = "blue", 
     main = "Scatterplot of Wind vs Pressure")

Scatterplot Matrices Example

Example: Draw scatterplot matrix of first 4 variables of the data frame iris (in the datasets package).

#>   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
#> 1          5.1         3.5          1.4         0.2  setosa
#> 2          4.9         3.0          1.4         0.2  setosa
#> 3          4.7         3.2          1.3         0.2  setosa
#> 4          4.6         3.1          1.5         0.2  setosa
#> 5          5.0         3.6          1.4         0.2  setosa
#> 6          5.4         3.9          1.7         0.4  setosa

# pch = 19 stands for points symbols type 19,
# which is solid circle
pairs(iris[,1:4], pch=19)

Linegraph Example – Draw a linegraph of the Data Science search counts vs Month.

#>   Month_Since_2004 DataScience MachineLearning
#> 1                1          12              16
#> 2                2          10              14
#> 3                3           7              12
#> 4                4           7              16
#> 5                5           5              14
#> 6                6           7              11

par(mfrow=c(1,2)) # 1 row, 2 columns (row-wise filling)
plot(ds$Month_Since_2004, ds$DataScience, type="l",
     main = "Line Plot for Data Science",
     xlab = "Number of Months Since 2004",
     ylab = "Data Science Search Counts") # type="l" represents lines

### Plot with points and line
plot(ds$Month_Since_2004, ds$DataScience, type="b", # type="b" represents both
     main = "Line and Point Plot for Data Science",
     xlab = "Number of Months Since 2004",
     ylab = "Data Science Search Counts") 

par(mfrow=c(1,1)) # reset to default

Multiple linegraphs Example

Example: We could place multiple linegraphs on one plot. Draw two linegraphs (DataScience vs Month_Since_2004, MachineLearning vs Month_Since_2004) on a single plot.

plot(ds$Month_Since_2004, ds$DataScience, type="l", col="red")
lines(ds$Month_Since_2004, ds$MachineLearning, col="blue")

Histogram Example

Example: Draw histogram of variable Sepal.Width in the data frame iris (in the datasets package).

data(iris)
hist(iris$Sepal.Width)

Histogram Matrices Example

library(Hmisc)
hist.data.frame(iris[,1:4])

Boxplot Example

Example:

1. Draw a boxplot of the variable wind in the data frame storms (in the tidyverse package).

2. Draw a boxplot of wind vs category of the data frame storms.

par(mfrow=c(2,2)) # Arrange the plots in 2*2 graphical matrix

### (a) boxplot of wind
boxplot(storms$wind, main = "(a) Vertical")
boxplot(storms$wind, horizontal = TRUE, main = "(a) Horizontal")
### (b) boxplot of wind over levels of category
### vertical boxplot by default
boxplot(wind ~ category, data = storms, main = "(b) Vertical")

### horizontal boxplot
boxplot(wind ~ category, data = storms, horizontal = TRUE, main = "(b) Horizontal")

par(mfrow=c(1,1))

Barplot Example

Example: In the storms dataset,

1. Draw a barplot of variable status.

2. Draw a barplot of variable of status and category.

### (a) Barplot
counts = table(storms$status)
barplot(counts)

counts
#> 
#>            disturbance          extratropical              hurricane 
#>                    171                   2151                   4803 
#>              other low subtropical depression      subtropical storm 
#>                   1453                    151                    298 
#>    tropical depression         tropical storm          tropical wave 
#>                   3569                   6830                    111

# Horizontal barplot
barplot(counts, horiz=TRUE)

### (b) Barplot of two variables
counts = table(storms$category, storms$status)
counts
#>    
#>     disturbance extratropical hurricane other low subtropical depression
#>   0           0             0         0         0                      0
#>   1           0             0      2548         0                      0
#>   2           0             0       993         0                      0
#>   3           0             0       593         0                      0
#>   4           0             0       553         0                      0
#>   5           0             0       116         0                      0
#>    
#>     subtropical storm tropical depression tropical storm tropical wave
#>   0                 0                   0              1             0
#>   1                 0                   0              0             0
#>   2                 0                   0              0             0
#>   3                 0                   0              0             0
#>   4                 0                   0              0             0
#>   5                 0                   0              0             0

barplot(counts, col= c("steelblue", "firebrick", "lightblue", "mistyrose", "lightcyan",
  "lavender", "cornsilk"), besid=TRUE, legend = rownames(counts),
  args.legend=list(title="Category"))

• category variable in storms represents the Saffir-Simpson storm category (range from 1 to 5), which the scale based only on a hurricane’s maximum sustained wind speed.

ggplot2

Why ggplot

1. Looks good!

• library(ggplot2) is used by fivethirtyeight, Financial Times, BBC, the Urban Institute, and more.

2. Flexible and expressive

• Constructing a visualization using library(ggplot2) is like playing a set of building blocks (it’s fun!)
• Add different layers of component parts to the plot.

3. Reproducible and scalable

• Once you have the plot-building-pipelines built, feed the data, select the code, Ctrl + Enter, then all done!
• It’s almost as easy to make the 100th chart as it is to make the 2nd chart.

4. Essential in analysis workflow

ggplot2– five important graphs – Summary

Assume we have the variables xvar, yvar in the data frame df.

### scatterplot
ggplot(data=df, mapping=aes(x=xvar, y=yvar)) +
  geom_point()

### linegraph
ggplot(data=df, mapping=aes(x=xvar, y=yvar)) +
  geom_line()

  #linegraph with dots
  ggplot(data=df, mapping=aes(x=xvar, y=yvar)) +
    geom_line() +
    geom_point()

### histogram
ggplot(data=df, mapping=aes(x=xvar)) +
  geom_histogram()

### boxplot
  # boxplot of one variable yvar
  ggplot(data=df, mapping=aes(y=yvar)) +
    geom_boxplot()
  # boxplot of the yvar vs xvar
  ggplot(data=df, mapping=aes(x=xvar, y=yvar)) +
    geom_boxplot()

### barplot of variable x
  # If not pre-counted, use geom_bar()
  ggplot(data=df, mapping=aes(x=xvar))+
   geom_bar()
  # If precounted, use geom_col()
  ggplot(data=df_counted, mapping=aes(x=xvar,y=counts))+
    geom_col()

ggplot2 – scatterplot – Example

Function: geom_point(x,y)

Example: Draw scatterplot of wind vs pressure of the data set storms (in the tidyverse package).

# Data preparation
library(tidyverse)
data(storms) # data frame in tidyverse package

storms[6923, "category"]=0 #Fixed one error value
storms$category = as.factor(storms$category)

# load the data set storms (in the tidyverse package)
ggplot(data = storms, mapping = aes(x=pressure, y=wind))+
  geom_point()

• Take a quick peek on dealing with overplotting

 # Changing the transparency
ggplot(storms, mapping = aes(x=pressure, y=wind)) + 
  geom_jitter(alpha=0.2)

ggplot2 – scatterplot matrices – Example

Function: ggpairs(df)

Example: Draw scatterplot matrix of first 4 variables of the data frame iris (in the datasets package).

data(iris)

library(GGally)
ggpairs(iris[,1:4])

ggplot2 – linegraph – Example

Function: geom_line()

Example: From https://trends.google.com/, we downloaded the time series data set GoogleTrendDataScience.csv with three variables Month_Since_2004, DataScience, MachineLearning.

1. Draw a linegraph of the Data Science search counts vs Month.

2. Draw a linegraph with points.

ds = read.csv("GoogleTrendDataScience.csv")

### Plot with line only
p1 = ggplot(ds, aes(x = Month_Since_2004, y = DataScience)) +
  geom_line()

### Plot with points and line
p2= ggplot(ds, aes(x = Month_Since_2004, y = DataScience)) +
  geom_line() +
  geom_point(size=1.5)

library(gridExtra)
grid.arrange(p1, p2, nrow = 1, ncol = 2)

ggplot2 – Multiple linegraphs – Example

Function: geom_line()

Example: We could place multiple linegraphs on one plot. Draw two linegraphs (DataScience vs Month_Since_2004, MachineLearning vs Month_Since_2004) on one plot.

ggplot(ds) +
  geom_line(aes(x=Month_Since_2004, y=DataScience), col = "tomato3", size = 1.5) +
  geom_line(aes(x=Month_Since_2004, y=MachineLearning), col="steelblue", size = 1.5)

ggplot2 – histogram – Example

Function: geom_histogram()

Example: Draw histogram of variable Sepal.Width in the data frame iris (in the datasets package).

## Illini orange and blue!
ggplot(iris, aes(x=Sepal.Width)) +
  geom_histogram(binwidth=0.2, color="#13294B", fill="#FF5F05")

ggplot2 – boxplot – Example

Functions: geom_boxplot()

Example:

1. Draw a boxplot of the variable wind in the data frame storms (in the tidyverse package).

2. Draw a boxplot of wind vs category of the data frame storms .

### (a) boxplot of wind
p1 = ggplot(storms, aes(y=wind))+
  geom_boxplot()



### (b) boxplot of wind over levels of category
### vertical boxplot by default
p2 = ggplot(storms, aes(x = category, y = wind))+
  geom_boxplot()

### horizontal boxplot
p3 = ggplot(storms, aes(y = category, x = wind))+
  geom_boxplot()
# Or
p4 = ggplot(storms, aes(x = category, y = wind))+
  geom_boxplot()+
  coord_flip()

grid.arrange(p1, p2, p3, p4, nrow = 2, ncol = 2)

ggplot2 – barplot – Example

Functions: geom_bar(), geom_col()

Example: In the storms dataset,

1. Draw a barplot of variable status.

2. Draw a barplot of variable of status and category

### Barplot
    # If not pre-counted, use geom_bar()
    ggplot(storms, aes(x = status))+
      geom_bar()+
      scale_x_discrete(guide = guide_axis(angle = 45)) ### Rotate the x axis label

    # If precounted, use geom_col()
    storms %>% 
      group_by(status) %>%
      summarise(count = n())%>%
      ggplot(aes(x = status, y = count)) +
      geom_col()+
      scale_x_discrete(guide = guide_axis(angle = 45)) ### Rotate the x axis label

# Horizontal barplot
ggplot(storms, aes(y=status))+
  geom_bar()

### Barplot of two variables
ggplot(storms, aes(x = status, fill = factor(category)))+
  geom_bar(position = "dodge")+
  scale_x_discrete(guide = guide_axis(angle = 45)) ### Rotate the x axis label

• category variable in storms represents the Saffir-Simpson storm category (range from 1 to 5), which the scale based only on a hurricane’s maximum sustained wind speed.

### Barplot
p1 = ggplot(storms, aes(x = status))+
  geom_bar()+
  scale_x_discrete(guide = guide_axis(angle = 45)) ### Rotate the x axis label

# Horizontal barplot
p2 = ggplot(storms, aes(y = status))+
  geom_bar()

### Barplot of two variables
p3 = ggplot(storms, aes(x = status, fill = factor(category))) +
  geom_bar(position = "dodge")+
  scale_x_discrete(guide = guide_axis(angle = 45)) ### Rotate the x axis label

grid.arrange(p1, p2, p3, layout_matrix = rbind(c(1,2), c(3,3)))

Want to save your ggplot?

ggplot(storms, aes(x = status, fill = factor(category))) +
  geom_bar(position = "dodge")+
  scale_x_discrete(guide = guide_axis(angle = 45))

ggsave("The Barplots.png") ### Save the last plot
ggsave("p1.png", plot = p1) ### Save the plot specified

Comments

• If a graph could be done with ggplot2, it could probably be done with base R graphics, and vice versa.

• To take a quick look at the data, the one-line base R graphics functions are quick and easy.

• For data visualization, ggplot2 is much easier to code and with a much better output.

Resources

• R Graph Gallery https://r-graph-gallery.com/

• Rstudio Cheatsheets https://www.rstudio.com/resources/cheatsheets/

• Reference Books (Free online books with excellent material)

  • R for Data Science: https://r4ds.had.co.nz/index.html
  • Hands-on Programming with R: https://rstudio-education.github.io/hopr/
  • R Graphics Cookbook: https://r-graphics.org/