• We will begin our journey into statistical graphics with the package ggplot2.
• This is a package by Hadley Wickham and is part of the tidyverse.
• It is a very comprehensive and easily adaptable language of graphics.

• A good place to start might be with what ggplot2 cannot do:
  • 3d graphs.
  • Interactive graphs, use ggvis
  • DAGs, see igraph
• We will now focus on all the things it can do.

• ggplot2 is built off the grammar of graphics with a very intuitive structure.
• The base graphics built into R require the use of many different functions and each of them seem to have their own method for how to use them.
  
• ggplot2 will be more fluid and the more you learn about it the more amazing of graphics you can create.

• We will get started with the components of every ggplot2 object:
  1. data
  2. aesthetic mappings between variables in the data and visual properties.
  3. At least one layer which describes how to render the data.
  4. Many of these are with the geom_foo() function.

• We will be working with data from fivethirtyeight
• We can access this by installing the package in R:

install.packages("fivethiryeight")
library(fivethirtyeight)

• We will work with data from the article: Comic Books Are Still Made By Men, For Men And About Men
• We can access this data by calling the following:

comic_characters

## # A tibble: 23,272 x 16
##    publisher page_id name  urlslug id    align eye   hair  sex   gsm  
##    <chr>       <int> <chr> <chr>   <chr> <ord> <chr> <chr> <chr> <chr>
##  1 Marvel       1678 Spid~ "\\/Sp~ Secr~ Good~ Haze~ Brow~ Male~ <NA> 
##  2 Marvel       7139 Capt~ "\\/Ca~ Publ~ Good~ Blue~ Whit~ Male~ <NA> 
##  3 Marvel      64786 "Wol~ "\\/Wo~ Publ~ <NA>  Blue~ Blac~ Male~ <NA> 
##  4 Marvel       1868 "Iro~ "\\/Ir~ Publ~ Good~ Blue~ Blac~ Male~ <NA> 
##  5 Marvel       2460 Thor~ "\\/Th~ No D~ Good~ Blue~ Blon~ Male~ <NA> 
##  6 Marvel       2458 Benj~ "\\/Be~ Publ~ Good~ Blue~ No H~ Male~ <NA> 
##  7 Marvel       2166 Reed~ "\\/Re~ Publ~ Good~ Brow~ Brow~ Male~ <NA> 
##  8 Marvel       1833 Hulk~ "\\/Hu~ Publ~ Good~ Brow~ Brow~ Male~ <NA> 
##  9 Marvel      29481 Scot~ "\\/Sc~ Publ~ <NA>  Brow~ Brow~ Male~ <NA> 
## 10 Marvel       1837 Jona~ "\\/Jo~ Publ~ Good~ Blue~ Blon~ Male~ <NA> 
## # ... with 23,262 more rows, and 6 more variables: alive <chr>,
## #   appearances <int>, first_appearance <chr>, month <chr>, year <int>,
## #   date <date>

• We can see the names of the variables by using the names() function:

names(comic_characters)

##  [1] "publisher"        "page_id"          "name"            
##  [4] "urlslug"          "id"               "align"           
##  [7] "eye"              "hair"             "sex"             
## [10] "gsm"              "alive"            "appearances"     
## [13] "first_appearance" "month"            "year"            
## [16] "date"

• We will begin with a basic graph of appearances by alignment

library(ggplot2)
ggplot(data=comic_characters, aes(x=align, y=appearances))

• We can see that all we have is the basic layout of axis.
• The data and aes gives us the basic layout.
• We need geom_foo() to make a proper graph.

• We can addgeom_point() to this:

ggplot(data=comic_characters, aes(x=align, y=appearances)) + 
  geom_point()

• This graph is not the best use for out data.
• We could try boxplots instead:

ggplot(data=comic_characters, aes(x=align, y=appearances)) + 
  geom_boxplot()

• We may be having a hard time seeing this due to the outliers in the data.
• We can try a log transform.
• The code below will do this
• By the end of today you will understand this code

comic_characters <- comic_characters %>%
  mutate(log_app = log(appearances))

• We could then consider simple bar graphs
• For example if we wanted to know how many characters of each type there were:

ggplot(data=comic_characters, aes(x=align)) + 
  geom_bar()

• Note: With bar graphs we only need the x-axis.

ggplot(data= <DATA>, aes(x=<X-VARIABLE>, y=<Y-VARIABLE>)) + 
    <GEOM_FUNCTION>()

• The basic aesthetics are mapping the data to the x and y axis.
• We can also add:
  • alpha: makes points transparent to see overlaps better
  • fill: Fills objects with color
  • color: Changes color of points or lines.
  • shape: Changes spape of points

ggplot(data=comic_characters, aes(x=align, y=log_app)) + 
  geom_point(aes(alpha=1/100))

• It can be hard to see the transparency when they are so close.
• We can set the transparency to a variable

• Set alpha=year
• How does this change things?

• We can easily change the color of points and lines using color

ggplot(data=comic_characters, aes(x=align, y=log_app)) + 
  geom_point(aes(color=publisher))

• Set color="blue"
• How does this change things?

• We can change the shape of points based on different variables.

ggplot(data=comic_characters, aes(x=align, y=log_app)) + 
  geom_point(aes(shape=publisher))

• Try using both shape and color.
• How does this add dimensionality to the graph?

• We can fill objects with color as well

ggplot(data=comic_characters, aes(x=align)) + 
  geom_bar(aes(fill="blue"))

• This doesnt have the same effect as color

• We can use other variables to add dimensionality

ggplot(data=comic_characters, aes(x=align)) + 
  geom_bar(aes(fill=publisher))

• There are many geom_foo() functions we can use.
• The Cheatsheet on ggplot() is a good place to start for more.

• We will discuss a concept that will help us greatly when it comes to working with our data.
• The usual way to perform multiple operations in one line is by nesting.

To consider an example we will look at the data provided in the gapminder package:

library(gapminder)
head(gapminder)

## # A tibble: 6 x 6
##   country     continent  year lifeExp      pop gdpPercap
##   <fct>       <fct>     <int>   <dbl>    <int>     <dbl>
## 1 Afghanistan Asia       1952    28.8  8425333      779.
## 2 Afghanistan Asia       1957    30.3  9240934      821.
## 3 Afghanistan Asia       1962    32.0 10267083      853.
## 4 Afghanistan Asia       1967    34.0 11537966      836.
## 5 Afghanistan Asia       1972    36.1 13079460      740.
## 6 Afghanistan Asia       1977    38.4 14880372      786.

• Let's say that we want to have the GDP per capita and life expectancy Kenya.
• Traditionally speaking we could do this in a nested manner:

filter(select(gapminder, country, lifeExp, gdpPercap), country=="Kenya")

• It is not easy to see exactly what this code was doing but we can write this in a manner that follows our logic much better.
• The code below represents how to do this with chaining.

gapminder %>%
    select(country, lifeExp, gdpPercap) %>%
    filter(country=="Kenya")

• We now have something that is much clearer to read.
• Here is what our chaining command says:
  1. Take the gapminder data
  2. Select the variables: country, lifeExp and gdpPercap.
  3. Only keep information from Kenya.
• The nested code says the same thing but it is hard to see what is going on if you have not been coding for very long.

• The result of this search is below:

# A tibble: 12 x 3
   country lifeExp gdpPercap
    <fctr>   <dbl>     <dbl>
 1   Kenya  42.270  853.5409
 2   Kenya  44.686  944.4383
 3   Kenya  47.949  896.9664
 4   Kenya  50.654 1056.7365
 5   Kenya  53.559 1222.3600
 6   Kenya  56.155 1267.6132
 7   Kenya  58.766 1348.2258
 8   Kenya  59.339 1361.9369
 9   Kenya  59.285 1341.9217
10   Kenya  54.407 1360.4850
11   Kenya  50.992 1287.5147
12   Kenya  54.110 1463.2493

• In the previous code we saw that we used %>% in the command you can think of this as saying then.
• For example:

gapminder %>%
    select(country, lifeExp, gdpPercap) %>%
    filter(country=="Kenya")

• This translates to:
  • Take Gapminder then select these columns select(country, lifeExp, gdpPercap) then filter out so we only keep Kenya

• We still might ask why we would want to do this.
• Chaining increases readability significantly when there are many commands.
• With many packages we can replace the need to perform nested arguments.
• The chaining operator is automatically imported from the magrittr package.

• Let's say that we wish to find the Euclidean distance between two vectors say, x1 and x2.
• We could use the math formula:

\[\sqrt{\sum(x_1-x_2)^2}\]

• In the nested manner this would be:

x1 <- 1:5; x2 <- 2:6
sqrt(sum((x1-x2)^2))

• However, if we chain this we can see how we would perform this mathematically.

# chaining method
(x1-x2)^2 %>% sum() %>% sqrt()

• If we did it by hand we would perform elementwise subtraction of x2 from x1 then we would sum those elementwise values then we would take the square root of the sum.

# chaining method
(x1-x2)^2 %>% sum() %>% sqrt()

## [1] 2.236068

• Many of us have been performing calculations by this type of method for years, so that chaining really is more natural for us.