2.1 Practice Data

In this session, we’ll use local public construction project records from:

• Construction of 2015 Lakeview Amphitheater (Source: Onondaga County)
• Renovation of 2018 Hancock International Airport (Source: Syracuse Regional Airport Authority)

You can find these datasets and documentation in this GitHub repository. These are the same data gathered by Legal Services of Central New York (LSCNY) and scraped, analyzed, and visualized for Legal Services’ and Urban Jobs Task Force’s collaborative publication: “Building Equity in the Construction Trades: A Racial Equity Impact Statement”.

2.2 Required Data & Packages

Run the following code in R to install and load packages required for this session.

• Package readr is used for importing and exporting data
• Package dplyr is built around a grammar of data manipulation, inspired by SQL
• Package scales allows one to easily format variables, e.g. percents and dollar amounts
• Package ggthemes, introduced last session, expands the ggplot2 library of preset themes

if(!require("readr")){install.packages("readr")}
if(!require("dplyr")){install.packages("dplyr")}
if(!require("scales")){install.packages("scales")}
if(!require("ggplot2")){install.packages("ggplot2")}
if(!require("ggthemes")){install.packages("ggthemes")}

library(readr)
library(dplyr)
library(scales)
library(ggplot2)
library(ggthemes)

Run the following code to read in (i.e. import) our practice data:

• hc contains all scraped records for Hancock renovations
• lv contains all scraped records for Lakeview renovations

url <- paste0("https://raw.githubusercontent.com/jamisoncrawford/",
              "wealth/master/Tidy%20Data/hancock_lakeview_tidy.csv")   # URL containing data

hc <- read_csv(file = url, 
               col_types = "ccDcccddddccliii") %>%     # Abbreviations specifying data classes
  filter(project == "Hancock")                         # Filters data for "Hancock" only

lv <- read_csv(file = url, 
               col_types = "ccDcccddddccliii") %>%     # Abbreviations specifying data classes
  filter(project == "Lakeview")                        # Filters data for "Lakeview" only

rm(url)                                                # Removes object `url`

ggplot(hc,                            # Call data layer
       aes(x = race, 
           y = gross)) +              # Map variables in aesthetics layer
  geom_jitter(width = 0.2, 
              alpha = 0.3,
              color = "tomato") +     # Specify geometry layer, with some attributes
  theme_light()                       # Unessential: A preset theme for clarity

3.1 On Coordinates

In general, the Coordinates layer follows a few rules:

• All functions begin with coord_, type this in-console to view functions via autocomplete
• Function coord_cartesian() is the most common and modifies a plot’s Cartesian plane

3.2 Flipping Coordinates

No, we’re not politely cursing our coordinates.

Instead, we swap the x- and y-axes, which is a best practice when using categorical variables. Why?

• With many levels, categorical variables can crowd one another.
• Laypersons tend to read charts from left to right, starting in the upper-left and working down.
• In other words, general audiences will read a chart like text, in a “Z” pattern.

We can use function ccord_flip() to achieve this. Note that x = and y= arguments are reversed.

ggplot(hc,
       aes(x = race, 
           y = gross)) +
  geom_jitter(width = 0.2, 
              alpha = 0.3,
              color = "tomato") +
  coord_flip() +                      # Swap x- and y-axes with `coord_flip()`
  theme_light()

3.3 Cartesian Planes

We can “zoom in” on a part of our plot with function coord_cartesian():

• Here, we’ll view two continuous variables, gross and net income
• Surprisingly, these can differ since net includes income from other projects
• We add a Loess regression curve to show association using geom_smooth()
  • LOESS: Locally estimated scatterplot smoothing (Wikipedia)
• First, view the plot without coord_cartesian()

ggplot(hc,
       aes(x = net,          # Note the substitution of `race` for `net`
           y = gross)) +
  geom_point() +             # We don't need jitter and use `geom_pont()`
  geom_smooth() +            # By default, this plots a Loess curve and standard error
  theme_light()

Now, we’ll use coord_cartesian() to specify where we’d like to “zoom in”.

• Here, we specify the minimum and maximum zoom limits with concatenate, or c()
• We set the x-axis range with xlim = and y-axis range with ylim =
• Note that “zooming in” does not eliminate the other data points!

ggplot(hc,
       aes(x = net,
           y = gross)) +
  geom_point() +
  geom_smooth() +
  coord_cartesian(xlim = c(0, 1000),       # We may specify limits for one or both axes
                  ylim = c(0, 2000)) +     # The range of limits in `c()` can differ
  theme_light()

3.4 Filtering Planes

While coord_cartesian() allows one to “zoom in”, limit functions will “filter” unseen data.

• Function xlim() with argument x = controls the x-axis limits
• Function ylim() with argument y = controls the y-axis limits
• Function lims() accepts both x = and y = arguments
• Remember to provided a minimum and maximum value in c()

ggplot(hc,
       aes(x = net,
           y = gross)) +
  geom_point() +
  geom_smooth() +
  lims(x = c(0, 1000),
       y = c(0, 2000)) +     # Alternatively: xlim(c(0, 1000)) + ylim(c(0, 2000))
  theme_light()

Questions. Why has the Loess curve changed? Why has the standard error changed?

3.5 Scales

Modifying the scales of the x- and y-axes requires you to follow a few rules, as well:

• All functions modifying scales being with scale_; explore with autocomplete
• You must specify the axis to modify, e.g. scale_x_ or scale_y_
• You must also specify the type of variable in the function call
  • For example, if categorical e.g., use scale_x_discrete()
  • Conversely, if continuous e.g., use scale_y_continuous()
• The following modifies x- and y-axes labels, or labels =

ggplot(hc,
       aes(x = net,
           y = gross)) +
  geom_point() +
  geom_smooth() +
  scale_x_continuous(name = "Unique Net Payments") +       # Specify continuous x-axis
  scale_y_continuous(name = "Unique Gross Payments") +     # Specify continuous y-axis
  theme_light()

Labels. Above, we’ve simply modifed the axes labels with argument name =.

We can use R package scales to easily format the data labels, which may:

• Add comma separators for values of 1,000 or more (label = "comma")
• Convert to dollars (labels = dollar) or percents (labels = percent)

library(scales)

ggplot(hc,
       aes(x = net,
           y = gross)) +
  geom_point() +
  geom_smooth() +
  scale_x_continuous(name = "Unique Net Payments",
                     labels = dollar) +             # `scales` makes formatting easy!
  scale_y_continuous(name = "Unique Gross Payments",
                     labels = dollar) +
  theme_light()

Breaks. We can also format breaks along each axis, i.e. labels every $500, or every $1K.

• These breaks are numeric and so do not require quotiation marks
• Specify breaks using the argument breaks =
• Individual breaks are concatenated in c()

library(scales)

ggplot(hc,
       aes(x = net,
           y = gross)) +
  geom_point() +
  geom_smooth() +
  scale_x_continuous(name = "Unique Net Payments",
                     labels = dollar, 
                     breaks = c(0, 500, 1000, 1500, 2000, 2500)) +     # Custom breaks
  scale_y_continuous(name = "Unique Gross Payments",
                     labels = dollar,
                     breaks = c(0, 500, 1000, 1500, 2000, 2500, 3000, 3500)) +
  theme_light()

Repeating Breaks. We don’t need to type out each “break” value by using package scales:

• Function pretty_breaks() from scales allows easy, incremental break formatting
• Argument n = specifies the numebr of equal breaks

library(scales)

ggplot(hc,
       aes(x = net,
           y = gross)) +
  geom_point() +
  geom_smooth() +
  scale_x_continuous(name = "Unique Net Payments",
                     labels = dollar, 
                     breaks = pretty_breaks(n = 6)) +     # Pretty breaks
  scale_y_continuous(name = "Unique Gross Payments",
                     labels = dollar,
                     breaks = pretty_breaks(n = 8)) +
  theme_light()

Custom Labels. We don’t have to use package scales, though it helps. Instead:

• We can create x- and y-axis labels that use non-data ink more parsimoniously
• This allows audiences to spend more time focusing on the data
• However, we need to provide sufficient context with some “data ink”
• The number of unique labels must match number of breaks!
• Here, by incuding “$”, we need not include “USD” in labels

ggplot(hc,
       aes(x = net,
           y = gross)) +
  geom_point() +
  geom_smooth() +
  scale_x_continuous(name = "Unique Net Payments",
                     labels = c("$ 0", "0.5", "1.0", "1.5", "2.0", "2.5 K"),
                     breaks = c(0, 500, 1000, 1500, 2000, 2500)) +
  scale_y_continuous(name = "Unique Gross Payments",
                     labels = c("$ 0", "0.5", "1.0", "1.5", "2.0", "2.5", "3 K"),
                     breaks = c(0, 500, 1000, 1500, 2000, 2500, 3000)) +
  theme_light()