Lecture Examples

Let’s examine choices that we can make with the examples from Week 1.

First we are going to load the R package “MXB017”².

library(MXB107)

and load the epa_data dataset.

data("epa_data")

Creating the Pareto Plot

To create the Pareto plot³.

The MXB107 package includes both the tidyverse package for manipulating data and the ggplot2 package, a very powerful package for producing very nice (aesthetically pleasing) plots. The syntax can be confusing, but for now, think of the ggplot2 commands as adding layers to a picture.

The simplest way to make a barplot is to just use geom_bar():

ggplot(epa_data,aes(make))+
  geom_bar(stat = "count")

But this is ugly, first of all there are:

length(unique(epa_data$make))
#> [1] 138

One hundred thirty-eight different makes represented in the data set; this makes the \(x\)-axis unreadable.

• What are some other problems with this chart?
• What if any information can you gather from this chart?
• What should you do to improve this chart and make it more informative?

Note that it probably makes more sense to look at only the major manufacturers. The first step is to create a data structure we will call make_data⁴ which will contain the top ten most common makes of cars in epa_data.

make_data<-epa_data%>%group_by(make)%>%
  summarise(count = n())%>%
  top_n(10)
#> Selecting by count

Now let’s see what the bar plot looks like

ggplot(make_data,aes(x = reorder(make,-count)))+
  geom_bar(aes(y = count),stat = "identity")

Now, modify the code below, adding appropriate labels for the axes, and a title to the plot

make_data<-epa_data%>%group_by(make)%>%
  summarise(count = n())%>%
  top_n(10)

ggplot(make_data,aes(x = reorder(make,-count)))+
  geom_bar(aes(y = count),stat = "identity")

make_data<-epa_data%>%group_by(make)%>%
  summarise(count = n())%>%
  top_n(10)

ggplot(make_data,aes(x = reorder(make,-count)))+
  geom_bar(aes(y = count),stat = "identity")+
  xlab("Manufacturer")+
  ylab("Frequency")+
  ggtitle("Manufacturers by Count")+
  theme(text = element_text(size = 10))

Creating the Line Plot

To create a line plot showing how EPA ratings for fuel economy while driving on the highway have changed we could simply use ggplot() to create a generic plot for epa_data with year for the \(x\)-axis and city for the \(y\)-axis. Then add a line with geom_line().

ggplot(epa_data,aes(x = year, y = city))+
  geom_line()

But these results are confusing and difficult to understand,

• What are some other problems with this chart?
• What if any information can you gather from this chart?
• What should you do to improve this chart and make it more informative?

Modify the code below using the ggplot function stat_summary to plot the average value of city for each year and add appropriate title, and labels for the axes⁵.

ggplot(epa_data,aes(x = year, y = city))

ggplot(epa_data,aes(x = year, y = city))+
  stat_summary(geom = "line", fun = "mean")+
  xlab("Year")+
  ylab("EPA City Fuel Economy (MPG)")+
  ggtitle("Yearly Average EPA City Fuel Economy (MPG) 1984-2021")
  theme(plot.title = element_text(hjust = 0.5))

Histograms

The line plot shows the trends over time in the average EPA City fuel economy ratings. Still, if we want to look more closely at a specific year to see the distribution or density of EPA City fuel economy ratings, we need to use a histogram.

ggplot(epa_data[epa_data$year==1990,],aes(city))+
  geom_histogram()

Let’s compare the histogram of city to hwy. To do this, we will need to pivot our table, so we have one variable that indicates whether the mileage is “city” or “highway”.

##  Create a new table with just the stuff we need.

df<-filter(epa_data,year==1990)%>%
  select(city,hwy)%>%
  pivot_longer(cols=c(city,hwy),names_to = "mileage")


ggplot(df,aes(x=value))+
geom_histogram(binwidth = 1)+facet_wrap(~mileage)

Is this a fair comparison? What principles of good graphical summaries does this violate?

• Identify what needs to be done to ensure that this plot conforms with the best principles for graphical summaries and make the changes.

• Adjust the bin width for the two charts and see how this changes your interpretation of the data.
  

df<-filter(epa_data,year==1990)%>%
  select(city,hwy)%>%
  pivot_longer(cols=c(city,hwy),names_to = "mileage")

ggplot(df,aes(x=value))+
geom_histogram(binwidth = 1)+facet_wrap(~mileage)

df<-filter(epa_data,year==1990)%>%
  select(city,hwy)%>%
  pivot_longer(cols=c(city,hwy),names_to = "mileage")

ggplot(df,aes(x=value))+
geom_histogram(aes(y=..density..),binwidth = 2)+
facet_wrap(~mileage)

Concluding Remarks

Graphical summaries may seem trivial and simplistic, but as you can see, it is easy to produce confusing or misleading graphical summaries; by adhering to the principles of good graphical summaries, you can avoid many problems.

Worksheet Practical Questions

# Answer here!

df<-filter(epa_data,year==1985)%>%
  pivot_longer(cols=c(city,hwy),names_to = "mileage")

ggplot(df,aes(x=value))+
geom_histogram(aes(y = ..density..),binwidth = 2)+
facet_wrap(~trans)

# Answer here!

ggplot(epa_data,aes(x = year, y = disp))+
stat_summary(fun = "mean", geom="line")+
xlab("Year")+
ylab("Average Engine Displacement (l)")+
ggtitle("Yearly Average Engine Displacement (l)")+
theme(plot.title=element_text(hjust=0.5))

# Answer here!

make_data<-epa_data%>%group_by(make)%>%
  summarise(mean_mpg = mean(city))%>%
  slice_max(order_by = mean_mpg,n = 10)

  ggplot(make_data,aes(x = reorder(make,-mean_mpg)))+
  geom_bar(aes(y = mean_mpg),stat = "identity")+
  xlab("Manufacturer")+
  ylab("Frequency")+
  ggtitle("Manufacturers by Count")