October 22, 2020

What we learned in the previous class

  • Selecting variables
  • Creating variables
  • Updating values
  • Subsetting tibbles/data frames
  • Aggregation

What we’ll learn today

  • Wrap up data processing
    • Warm-up exercise
    • Pivoting data with gather() and spread()
  • Overview of ggplot2
  • Build bar plots + exercise
  • Build other plots + exercises

Wrap up data processing

Let’s begin by reading our data

library(tidyverse)
covid <- read_csv('https://rb.gy/lzlylj')
crime <- read_csv('https://rb.gy/5zuayh')

Exercise - 10 minutes

  • From the covid dataset, determine the location-month combination with the highest number of estimated infections per capita. Exclude dates prior to July and 'Global' values in location. To do this:
    • Filter rows based on the stated date and locaton conditions
    • Create a new variable called infections_per_capita that reports est_infections over population
    • Group by location and month and sum infections_per_capita
    • With arrange, sort by infections_per_capita so that rows are in descending (desc()) order
covid %>% 
  filter(
    month >= 7 & 
    ! location %in% 'Global'
  )

Exercise - 10 minutes

  • From the covid dataset, determine the location-month combination with the highest number of estimated infections per capita. Exclude dates prior to July and 'Global' values in location. To do this:
    • Filter rows based on the stated date and locaton conditions
    • Create a new variable called infections_per_capita that reports est_infections over population
    • Group by location and month and sum infections_per_capita
    • With arrange, sort by infections_per_capita so that rows are in descending (desc()) order
covid %>% 
  filter(
    month >= 7 & 
    ! location %in% 'Global'
  ) %>% 
  mutate(infections_per_capita = est_infections / population)

Exercise - 10 minutes

  • From the covid dataset, determine the location-month combination with the highest number of estimated infections per capita. Exclude dates prior to July and 'Global' values in location. To do this:
    • Filter rows based on the stated date and locaton conditions
    • Create a new variable called infections_per_capita that reports est_infections over population
    • Group by location and month and sum infections_per_capita
    • With arrange, sort by infections_per_capita so that rows are in descending (desc()) order
covid %>% 
  filter(
    month >= 7 & 
    ! location %in% 'Global'
  ) %>% 
  mutate(infections_per_capita = est_infections / population) %>% 
  group_by(location, month) %>% 
  summarise(infections_per_capita = sum(infections_per_capita, na.rm = TRUE))

Exercise - 10 minutes

covid %>% 
  filter(
    month >= 7 & 
    ! location %in% 'Global'
  ) %>% 
  mutate(infections_per_capita = est_infections / population) %>% 
  group_by(location, month) %>% 
  summarise(infections_per_capita = sum(infections_per_capita, na.rm = TRUE)) %>% 
  arrange(desc(infections_per_capita)) %>% 
  head()
## # A tibble: 6 x 3
## # Groups:   location [4]
##   location    month infections_per_capita
##   <chr>       <dbl>                 <dbl>
## 1 Manipur        12                 0.167
## 2 Meghalaya      12                 0.147
## 3 Malta          11                 0.135
## 4 Malta          12                 0.127
## 5 West Bengal    12                 0.107
## 6 Manipur        11                 0.106

Reshape tibbles

  • gather() and spread() are the functions we use to reshape (or pivot) data
  • Reshape means to convert…
    • Columns into rows (with gather())
    • Rows into columns (with spread())

Reshape tibbles

  • gather() and spread() are the functions we use to reshape data
  • Reshape means to convert…
    • Columns into rows (with gather())
    • Rows into columns (with spread())

Reshape tibbles

  • gather()
  • Convert columns into rows
  • Arguments
    • Tibble
    • Name of column for column names
    • Name of column for values in columns
    • Which columns you want turn into rows
crime %>% 
  select(report_number, occurred_date, reported_date) %>% 
  gather(year_type, year, c(occurred_date, reported_date))

Reshape tibbles

  • gather()
  • Convert columns into rows
  • Arguments
    • Tibble
    • Name of column for column names
    • Name of column for values in columns
    • Which columns you want turn into rows
crime %>% 
  select(report_number, occurred_date, reported_date) %>% 
  gather(year_type, year, 2:3)

Reshape tibbles

crime %>% 
  tail(2) %>% 
  select(report_number, occurred_date, reported_date) %>% 
  gather(year_type, year, c(occurred_date, reported_date))
## # A tibble: 4 x 3
##   report_number year_type     year      
##   <chr>         <chr>         <date>    
## 1 2010-044880   occurred_date 2010-02-06
## 2 2020-218244   occurred_date 2020-07-20
## 3 2010-044880   reported_date 2010-02-08
## 4 2020-218244   reported_date 2020-07-20

Reshape tibbles

  • spread()
  • Convert rows into columns
  • Arguments
    • Tibble
    • Name of column that contains the names of the new columns
    • Name of columns that contains the values of the new columns
covid %>% 
  filter(! mobility_data_type %in% NA & location %in% c('Peru', 'Ecuador', 'Colombia')) %>% 
  group_by(mobility_data_type, location) %>% 
  summarise(mobility_composite = mean(mobility_composite, na.rm = TRUE)) %>% 
  spread(location, mobility_composite)

Reshape tibbles

covid %>% 
  filter(! mobility_data_type %in% NA & location %in% c('Peru', 'Ecuador', 'Colombia')) %>% 
  group_by(mobility_data_type, location) %>% 
  summarise(mobility_composite = mean(mobility_composite, na.rm = TRUE)) %>% 
  spread(location, mobility_composite)
## # A tibble: 2 x 4
## # Groups:   mobility_data_type [2]
##   mobility_data_type Colombia Ecuador  Peru
##   <chr>                 <dbl>   <dbl> <dbl>
## 1 observed              -38.7   -51.1 -54.6
## 2 projected             -12.8   -38.7 -44.0

Reshape tibbles

Exercise #1 - 3 minutes

  • With crime, use gather() to convert precinct, beat, and neighborhood columns into rows
    • Name the two new columns whatever you would like
    • Keep only the columns you created and report_number

Exercise #2 - 5 minutes

  • With crime, report the number of incidents by offense_parent_group and precinct. To do this…
    • First, count incidents with group_by() and summarise()
    • Second, use spread() to turn precinct into columns with counts of incidents

Exercise #1 - 3 minutes

  • With crime, use gather() to convert precinct, beat, and neighborhood columns into rows
    • Name the two new columns whatever you would like
    • Keep only the columns you created and report_number
crime %>% 
  gather(group, value, c(precinct, beat, neighborhood)) %>%

Exercise #1 - 3 minutes

  • With crime, use gather() to convert precinct, beat, and neighborhood columns into rows
    • Name the two new columns whatever you would like
    • Keep only the columns you created and report_number
crime %>% 
  gather(group, value, c(precinct, beat, neighborhood)) %>%
  select(report_number, group, value) %>%
  sample_n(8)
## # A tibble: 8 x 3
##   report_number group        value          
##   <chr>         <chr>        <chr>          
## 1 2019-119558   neighborhood <NA>           
## 2 2010-214293   neighborhood ALASKA JUNCTION
## 3 2020-236131   precinct     South          
## 4 2012-903035   neighborhood GREENWOOD      
## 5 2012-099379   precinct     UNKNOWN        
## 6 2015-150623   beat         R3             
## 7 2012-397750   precinct     North          
## 8 2015-240127   precinct     West

Exercise #2 - 5 minutes

  • With crime, report the number of incidents by offense_parent_group and precinct. To do this…
    • First, count incidents with group_by() and summarise()
    • Second, use spread() to turn precinct into columns with counts of incidents
crime %>% 
  filter(! precinct %in% c('UNKNOWN', NA)) %>% 
  group_by(precinct, offense_parent_group) %>% 
  summarise(n = n())

Exercise #2 - 5 minutes

  • With crime, report the number of incidents by offense_parent_group and precinct. To do this…
    • First, count incidents with group_by() and summarise()
    • Second, use spread() to turn precinct into columns with counts of incidents
crime %>% 
  filter(! precinct %in% c('UNKNOWN', NA)) %>% 
  group_by(precinct, offense_parent_group) %>% 
  summarise(n = n()) %>% 
  spread(precinct, n)
## # A tibble: 32 x 6
##    offense_parent_group                      East North South Southwest  West
##    <chr>                                    <int> <int> <int>     <int> <int>
##  1 ANIMAL CRUELTY                               1     3     1        NA     3
##  2 ARSON                                       26    67    33        19    33
##  3 ASSAULT OFFENSES                          2979  4488  2887      1882  4777
##  4 BAD CHECKS                                 146   321   135        90   213
##  5 BRIBERY                                      1     2     1        NA    NA
##  6 BURGLARY/BREAKING&ENTERING                2054  4898  2135      1396  2618
##  7 COUNTERFEITING/FORGERY                     114   261    89        59   165
##  8 CURFEW/LOITERING/VAGRANCY VIOLATIONS        12     8     8         4    91
##  9 DESTRUCTION/DAMAGE/VANDALISM OF PROPERTY  1845  3489  1871      1355  2459
## 10 DRIVING UNDER THE INFLUENCE                405   663   357       256   564
## # … with 22 more rows

Visualization with ggplot2

ggplot2

Plot types

Plot types

Plot types

Plot types

What we will learn

Requests to learn another plot type?

Plot types

1 discrete variable (plus other optional discrete and/or continuous variables)

1 continuous variable

Plot types

2 continuous variables

1 continuous variable + date variable

Let’s make sure you’re able to plot your data

library(tidyverse)
mpg %>% ggplot() + geom_point(aes(displ, hwy))

Let’s make sure you’re able to plot your data

  • If you see a plot, you’re ready to go!
  • If you do not, reinstall tidyverse and re-run the test code
install.packages('tidyverse')
library(tidyverse)
mpg %>% ggplot(aes(displ, hwy))) + geom_point()
  • If it still didn’t work, install ggplot2 and re-run the test code
install.packages('ggplot2')
library(ggplot2)
mpg %>% ggplot(aes(displ, hwy)) + geom_point()

How to plot data

  1. Pick your data
  2. Pick your chart type (Geom)
  3. Program stuff like chart titles, axis names, axis types, and legends (Scales)
  4. Make it pretty (Themes)
  5. Keep tuning and reuse code

ggplot basics

Geoms

  • Every plot needs a ggplot() function and a geom layer
  • Separate layers and other plotting functions with +
ggplot() + geom_bar()       # create bar and stacked bar plots
ggplot() + geom_histogram() # create histograms
ggplot() + geom_point()     # create scatter plots
ggplot() + geom_line()      # create line plots
  • Your data is an argument in the ggplot() function
covid %>% ggplot() + geom_bar()
ggplot(covid) + geom_bar()

Geoms

  • Map your data to the plot with aes()
    • aes() indicates the variables that affect the chart aesthetics
    • aes() can be an argument in ggplot() or the geom
    • Many arguments in the aes() function
    • aes() arguments reference variables in your tibble
covid %>% ggplot(aes(mobility_data_type)) + geom_bar()
covid %>% ggplot() + geom_bar(aes(mobility_data_type))
aes(x = NULL, y = NULL, color = NULL
  , fill = NULL, alpha = NULL, label = NULL
  , shape = NULL, size = NULL, group = NULL
  , linetype = NULL
  )

Geoms

covid %>% ggplot(aes(x = mobility_data_type)) + geom_bar()
covid %>% ggplot() + geom_bar(aes(x = mobility_data_type))

Geoms

covid %>% ggplot(aes(x = mobility_data_type, fill = mobility_data_type)) + geom_bar()

Geoms

covid %>% ggplot(aes(x = mobility_data_type, fill = as.character(month))) + geom_bar()

Exercise - 7 minutes

  • Build a bar chart with crime data that shows how many incidents occurred in beat
  • Only include these beat values: c('B3', 'E3', 'D2', 'R2', 'O1', 'C3', 'K3')
  • Color the bars using offense
  • Only include these offense values: c('Impersonation', 'Driving Under the Influence', 'Pocket-picking', 'Embezzlement')

Hint

# Use to include/exclude values
filter()

Exercise - 7 minutes

  • Build a bar chart with crime data that shows how many incidents occurred in beat
  • Only include these beat values: c('B3', 'E3', 'D2', 'R2', 'O1', 'C3', 'K3')
  • Color the bars using offense
  • Only include these offense values: c('Impersonation', 'Driving Under the Influence', 'Pocket-picking', 'Embezzlement')

Hint

crime %>%
  filter() %>%
  ggplot()

Exercise - 7 minutes

crime %>% 
  filter(
    beat %in% c('B3', 'E3', 'D2', 'R2', 'O1', 'C3', 'K3') &
    offense %in% c('Impersonation', 'Driving Under the Influence'
                  , 'Pocket-picking', 'Embezzlement')
    ) %>%
  ggplot(aes(beat, fill = offense)) + 
  geom_bar()

Scales

Axis, legend, and chart titles

Axis names

  • Keep names short-ish
  • Consider including units in axis name
  • It is possible you don’t need an axis name

2 options

# Option 1
labs(x = 'X Axis Title', y = 'Y Axis Title')

# Option 2
xlab('X Axis Title')
ylab('Y Axis Title')

Axis, legend, and chart titles

covid %>% 
  filter(! mobility_data_type %in% NA) %>% 
  ggplot(aes(x = mobility_data_type, fill = mobility_data_type)) + 
  geom_bar() + 
  labs(x = 'Mobility Data Type', y = 'Observations (#)')

Axis, legend, and chart titles

Try adding axis names to the beat plot you made during the exercise

crime %>% 
  filter(
    beat %in% c('B3', 'E3', 'D2', 'R2', 'O1', 'C3', 'K3') &
    offense %in% c('Impersonation', 'Driving Under the Influence'
                  , 'Pocket-picking', 'Embezzlement')
    ) %>%
  ggplot(aes(beat, fill = offense)) + 
  geom_bar() + 
  labs(x = 'Seattle Beats', y = 'Incidents by Offense\nClearance Group (#)')

Axis, legend, and chart titles

Legend names

  • Keep names short
  • Be mindful of what the legend is for
    • Color? Fill? Size? etc.
  • Consider hiding the legend title for a clean look
labs(fill = '')
labs(fill = element_blank())
labs(fill = NULL)
labs(colour = 'Check out these colors')

Axis, legend, and chart titles

covid %>% 
  filter(! mobility_data_type %in% NA) %>% 
  ggplot(aes(x = mobility_data_type, fill = mobility_data_type)) + 
  geom_bar() + 
  labs(x = 'Mobility Data Type', y = 'Observations (#)', fill = 'Check out these colors')

Axis, legend, and chart titles

Try removing the legend name from your beat plot

Axis, legend, and chart titles

Try removing the legend name from your beat plot

crime %>% 
  filter(
    beat %in% c('B3', 'E3', 'D2', 'R2', 'O1', 'C3', 'K3') &
    offense %in% c('Impersonation', 'Driving Under the Influence'
                  , 'Pocket-picking', 'Embezzlement')
    ) %>%
  ggplot(aes(beat, fill = offense)) + 
  geom_bar() + 
  labs(
    x = 'Seattle Beats'
    , y = 'Incidents by Offense\nClearance Group (#)'
    , fill = element_blank()
    )

Axis, legend, and chart titles

Chart titles and subtitles

  • Consider making titles (or subtitles) descriptive
  • Don’t over use subtitles
# Option 1
labs(title = NULL, subtitle = NULL)

# Option 2
ggtitle(title = NULL, subtitle = NULL)

Axis, legend, and chart titles

covid %>% 
  filter(! mobility_data_type %in% NA) %>% 
  ggplot(aes(x = mobility_data_type, fill = mobility_data_type)) + 
  geom_bar() + 
  labs(
    x          = 'Mobility Data Type'
    , y        = 'Observations (#)'
    , fill     = element_blank()
    , title    = '2020 Covid-19 Mobility Types'
    , subtitle = 'Nearly two-thirds of mobility data are observed'
    )

Axis, legend, and chart titles

Axis, legend, and chart titles

Try adding a title to the beat plot

Axis, legend, and chart titles

Try adding a title to the beat plot

crime %>% 
  filter(
    beat %in% c('B3', 'E3', 'D2', 'R2', 'O1', 'C3', 'K3') &
    offense %in% c('Impersonation', 'Driving Under the Influence'
                  , 'Pocket-picking', 'Embezzlement')
    ) %>%
  ggplot(aes(beat, fill = offense)) + 
  geom_bar() + 
  labs(
  x = 'Seattle Beats'
  , y = 'Incidents by Offense\nClearance Group (#)'
  , fill = element_blank()
  , title = 'An Amazing Title'
  )

Axis units

  • Change axis units
  • This is a good idea and you should do it when its relevant
  • Good for x and y axes
install.packages('scales')
library(scales)

scale_y_continuous(labels = percent) # Add a percentage sign to numbers on axis
scale_y_continuous(labels = dollar)  # Add a dollar sign to numbers on axis
scale_y_continuous(labels = comma)   # Add a comma to numbers on axis

Axis units

covid %>% 
  filter(! mobility_data_type %in% NA) %>% 
  ggplot(aes(x = mobility_data_type, fill = mobility_data_type)) + 
  geom_bar() + 
  labs(
    x          = 'Mobility Data Type'
    , y        = 'Observations (#)'
    , fill     = element_blank()
    , title    = '2020 Covid-19 Mobility Types'
    , subtitle = 'Nearly two-thirds of mobility data are observed'
    ) + 
  scale_y_continuous(labels = comma)

Axis units

Themes

Themes

  • Change layer and background colors
  • Change fonts
  • Change plot borders/boundaries and ticks
  • Pre-built themes vs. custom themes
    • ggplot2 themes
    • ggthemes themes
install.packages('ggthemes')
library(ggthemes)

Pre-built themes

Selected ggplot2 themes

theme_classic()
theme_minimal()
theme_dark()

Selected ggthemes themes

theme_stata() + scale_colour_stata() # scale_fill_stata()
theme_economist() + scale_colour_economist() # scale_fill_economist()
theme_fivethirtyeight() + scale_color_fivethirtyeight() # scale_fill_fivethirtyeight()
theme_wsj() + scale_colour_wsj() # scale_fill_wsj()
theme_pander() + scale_colour_pander() # scale_fill_pander()
theme_hc(bgcolor = "darkunica") + scale_colour_hc("darkunica") # scale_fill_hc("darkunica")

Pre-built themes

covid %>% 
  filter(! mobility_data_type %in% NA) %>% 
  ggplot(aes(x = mobility_data_type, fill = mobility_data_type)) + 
  geom_bar() + 
  labs(
    x          = 'Mobility Data Type'
    , y        = 'Observations (#)'
    , fill     = element_blank()
    , title    = '2020 Covid-19 Mobility Types'
    , subtitle = 'Nearly two-thirds of mobility data are observed'
    ) + 
  scale_y_continuous(labels = comma) +
  theme_economist() + 
  scale_fill_economist()

Pre-built themes

Economist theme

Pre-built themes

Pre-built themes

Try adding a theme to the beat plot

Custom themes

  • You can create custom themes with theme()
  • Greater control over chart aesthestics
  • Use custom themes in place of pre-built themes if pre-built themes don’t cut it
  • Also, use custom themes to augment pre-built themes

Custom themes

theme(
  plot.title         = element_text(size=14, face="bold", vjust=1)
  , plot.background  = element_blank()
  , panel.grid.major = element_blank()
  , panel.grid.minor = element_blank()
  , panel.border     = element_blank()
  , panel.background = element_blank()
  , axis.ticks       = element_blank()
  , axis.text        = element_text(colour="black", size=12)
  , axis.text.x      = element_text(angle=45, hjust=1)
  , legend.title     = element_blank()
  , legend.position  = "none"
  , legend.text      = element_text(size=12)
  )

Exercise - 5 minutes

We answered the two questions below in the previous class. How would you turn the tabular output data into a chart?

  • In covid, which location reports the largest number of estimated infections in September, excluding 'Global' values?
  • In crime, of all 'ASSAULT OFFENSES' values in offense_parent_group which offense value has the fewest incidents?
covid %>% 
  group_by(month, location) %>% 
  summarise(est_infections = sum(est_infections, na.rm = TRUE)) %>%
  filter(month %in% 9 & ! location %in% 'Global') %>% 
  arrange(desc(est_infections)) %>%
  head()

Exercise - 5 minutes

How would you turn the tabular output data into a chart?

covid %>% 
  group_by(month, location) %>% 
  summarise(est_infections = sum(est_infections, na.rm = TRUE)) %>%
  filter(month %in% 9 & ! location %in% 'Global') %>% 
  arrange(desc(est_infections)) %>%
  head() %>% 
  ggplot(aes(x = reorder(location, est_infections), y = est_infections, fill = location)) + 
  geom_bar(stat = "identity") + 
  guides(fill=guide_legend(nrow=3)) + 
  labs(x = 'Location', y = 'Estimated Infections (#)') + 
  coord_flip() + # coord_flip() is a new function
  ggthemes::theme_hc()

Exercise - 5 minutes

How would you turn the tabular output data into a chart?

Exercise - 5 minutes

How would you turn the tabular output data into a chart?

covid %>% 
  group_by(month, location) %>% 
  summarise(est_infections = sum(est_infections, na.rm = TRUE)) %>%
  filter(month %in% 9 & ! location %in% 'Global') %>% 
  arrange(desc(est_infections)) %>%
  head() %>% 
  ggplot(aes(x = reorder(location, est_infections), y = est_infections, fill = location)) + 
  geom_bar(stat = "identity") + 
  guides(fill=guide_legend(nrow=3)) + 
  labs(x = 'Location', y = 'Estimated Infections (#)') + 
  coord_flip() + # coord_flip() is a new function
  ggthemes::theme_hc()

New argument and functions

geom_bar(stat = 'identity')       # IMPORTANT: use stat when working with aggregated data
reorder(location, est_infections) # reorder() to sort rows/columns in your visualization
coord_flip()                      # coord_flip(): x becomes y; y becomes x

Exercise - 5 minutes

We answered the two questions below in the previous class. How would you turn the tabular output data into a chart?

  • In covid, which location reports the largest number of estimated infections in September, excluding 'Global' values?
  • In crime, of all 'ASSAULT OFFENSES' values in offense_parent_group which offense value has the fewest incidents?
crime %>% 
  group_by(offense_parent_group, offense) %>% 
  summarise(incident = n()) %>%
  filter(offense_parent_group %in% 'ASSAULT OFFENSES' & ! offense %in% NA)
## # A tibble: 3 x 3
## # Groups:   offense_parent_group [1]
##   offense_parent_group offense            incident
##   <chr>                <chr>                 <int>
## 1 ASSAULT OFFENSES     Aggravated Assault     3754
## 2 ASSAULT OFFENSES     Intimidation           3963
## 3 ASSAULT OFFENSES     Simple Assault         9527

Exercise - 5 minutes

How would you turn the tabular output data into a chart?

crime %>% 
  group_by(offense_parent_group, offense) %>% 
  summarise(incident = n()) %>%
  filter(offense_parent_group %in% 'ASSAULT OFFENSES' & ! offense %in% NA) %>% 
  ggplot(aes(reorder(offense, -incident), incident)) +
  geom_bar(stat = 'identity') + 
  theme_classic()

Exercise - 15 minutes

  • From the crime dataset, report the mean difference in reported_date and occurred_date values by beat.
    • The variable that tells us the difference between reported_date and occurred_date should be called date_dif
    • In your output of mean values by beat only report rows where the number of beat values in the dataset is greater than 3000
    • You’ll use transmute and filter as well as group_by and summarise to solve this problem
    • The dataframe should have three variables: beat, date_dif, n
  • Create a bar plot that that shows date_dif values by beat

Hint

crime %>% 
  transmute(
    beat
    , date_dif = reported_date - occurred_date
  ) %>%

Exercise - 15 minutes

  • From the crime dataset, report the mean difference in reported_date and occurred_date values by beat.
    • The variable that tells us the difference between reported_date and occurred_date should be called date_dif
    • In your output of mean values by beat only report rows where the number of beat values in the dataset is greater than 3000
    • You’ll use transmute and filter as well as group_by and summarise to solve this problem
    • The dataframe should have three variables: beat, date_dif, n
  • Create a bar plot that that shows date_dif values by beat
crime %>% 
  transmute(
    beat
    , date_dif = reported_date - occurred_date
  ) %>% 
  group_by(beat) %>% 
  summarise(
    date_dif = mean(date_dif, na.rm = TRUE)
    , n = n()
  ) %>%

Exercise - 15 minutes

  • From the crime dataset, report the mean difference in reported_date and occurred_date values by beat.
    • The variable that tells us the difference between reported_date and occurred_date should be called date_dif
    • In your output of mean values by beat only report rows where the number of beat values in the dataset is greater than 3000
    • You’ll use transmute and filter as well as group_by and summarise to solve this problem
    • The dataframe should have three variables: beat, date_dif, n
  • Create a bar plot that that shows date_dif values by beat
crime %>% 
  transmute(
    beat
    , date_dif = reported_date - occurred_date
  ) %>% 
  group_by(beat) %>% 
  summarise(
    date_dif = mean(date_dif, na.rm = TRUE)
    , n = n()
  ) %>% 
  filter(n > 3000) %>%

Exercise - 15 minutes

crime %>% 
  transmute(
    beat
    , date_dif = reported_date - occurred_date
  ) %>% 
  group_by(beat) %>% 
  summarise(
    date_dif = mean(date_dif, na.rm = TRUE)
    , n = n()
  ) %>% 
  filter(n > 3000) %>% 
  ggplot(aes(x = beat, y = date_dif)) +
  geom_bar(stat = 'identity')

Exercise - 15 minutes

Exercise - 5 minutes

  • Take an extra 5 minutes to do the following
    • Add a theme to your plot
    • Change the axis names
    • Add a plot title
    • Arrange the beat columns by date_dif
    • Remove NA beat values

Exercise - 5 minutes

Exercise - 5 minutes

crime %>% 
  transmute(
    beat
    , date_dif = reported_date - occurred_date
  ) %>% 
  group_by(beat) %>% 
  summarise(
    date_dif = mean(date_dif, na.rm = TRUE)
    , n = n()
  ) %>% 
  filter(n > 3000 & ! beat %in% c(NA, 'UNKNOWN')) %>% 
  ggplot(aes(reorder(beat, -date_dif), date_dif)) +
  geom_bar(stat = 'identity') + 
  labs(
    y = 'Average delay in reporting (days)'
    , x = 'Beat'
    , title = 'Beat U3 sees the longest delay in events reporting'
    ) + 
  theme_classic()

Histograms, scatterplots, and line plots

Histograms: 1 continuous variable

  • bins is the argument you use to indicate the number of columns in your histogram
covid %>% 
  ggplot(aes(mobility_composite)) + 
  geom_histogram()

Histograms: 1 continuous variable

  • bins is the argument you use to indicate the number of columns in your histogram
covid %>% 
  ggplot(aes(mobility_composite)) + 
  geom_histogram(bins = 10)

Scatterplots: 2 continuous variables

covid %>% 
  ggplot(aes(x = deaths_p100k, y = est_infections_p100k)) + 
  geom_point()

Scatterplots: 2 continuous variables

covid %>% 
  ggplot(aes(x = deaths_p100k, y = est_infections_p100k)) + 
  geom_point()
geom_count() # adds a size component to your scatterplot
geom_jitter() # adjusts cartesian location of point relative to other points

Line plot: 1 continuous variable + date variable

covid %>% 
  group_by(date) %>%
  summarise(total_tests = sum(total_tests, na.rm = TRUE)) %>% 
  ggplot(aes(x = date, y = total_tests)) +
  geom_line(stat = 'identity')

Line plot: 1 continuous variable + date variable

covid %>% 
  group_by(date) %>%
  summarise(total_tests = sum(total_tests, na.rm = TRUE)) %>% 
  ggplot(aes(x = date, y = cumsum(total_tests))) + # cumsum() is a new function!
  geom_line(stat = 'identity')

Exercise - 5 minutes

  • With either dataset, covid or crime, create a scatterplot or line plot
  • Include one aes() arguments in addition to x and y
    • For example, size, shape, color or alpha
  • Line plots: 1 continuous variable + date variable
  • Scatterplot: 2 continuous variables
  • Use group_by() and summarise() if you would like
 crime %>% 
  filter(! precinct %in% NA) %>%
  ggplot(aes(reported_date, occurred_date, color = precinct)) + 
  geom_point() + 
  theme_classic()

Exercise - 5 minutes

 crime %>% 
  filter(! precinct %in% NA) %>%
  ggplot(aes(reported_date, occurred_date, color = precinct)) + 
  geom_point() + 
  theme_classic()

Exercise - 15 minutes

  • Create a plot from the crime dataset that shows the number of incidents by a character variable (you pick) and a time of day variable that indicates if the incident occurred in the AM or PM (first half or second half of the day)
    • Create a new variable with mutate() called occurred_time_ampm that tells whether an incident occurred in the AM or PM.
    • Make sure your axes are labeled correctly
    • Give your chart a title
    • Use occurred_time_ampm in the color or fill arguments in aes()
    • Use a pre-built theme

Exercise - 15 minutes

Exercise - 15 minutes

crime %>% 
  mutate(occurred_time_ampm = ifelse(occurred_time >= 1200, 'PM', 'AM')) %>% 
  filter(
    ! occurred_time_ampm %in% NA & 
    ! precinct %in% NA
    ) %>%
  group_by(precinct, occurred_time_ampm) %>%
  summarise(n = n()) %>%
  ungroup %>%
  ggplot(aes(reorder(precinct, n), n, fill = occurred_time_ampm)) + 
  geom_bar(stat = 'identity') + 
  coord_flip() + 
  theme_wsj() + 
  scale_fill_wsj() + 
  labs(title = 'Most incidents occur\nin the PM', fill = 'Time of Day')