Data wrangling - II

Lecture 4

Dr. Greg Chism

University of Arizona
INFO 526

Setup

# load packages
library(countdown)
library(tidyverse)
library(glue)
library(scales)
library(ggthemes)

# set theme for ggplot2
ggplot2::theme_set(ggplot2::theme_minimal(base_size = 14))

# set width of code output
options(width = 65)

# set figure parameters for knitr
knitr::opts_chunk$set(
  fig.width = 7,        # 7" width
  fig.asp = 0.618,      # the golden ratio
  fig.retina = 3,       # dpi multiplier for displaying HTML output on retina
  fig.align = "center", # center align figures
  dpi = 300             # higher dpi, sharper image
)

Transforming and reshaping a single data frame (cont.)

Data: Hotel bookings

• Data from two hotels: one resort and one city hotel
• Observations: Each row represents a hotel booking

hotels <- read_csv("https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2020/2020-02-11/hotels.csv")

Scenario 1

We…

have a single data frame

want to slice it, and dice it, and juice it, and process it, so we can plot it

Monthly bookings

Come up with a plan for making the following visualization and write the pseudocode.

Livecoding

Reveal below for code developed during live coding session.

hotels <- hotels |>
  mutate(
    arrival_date_month = fct_relevel(arrival_date_month, month.name),
    season = case_when(
      arrival_date_month %in% c("December", "January", "February") ~ "Winter",
      arrival_date_month %in% c("March", "April", "May") ~ "Spring",
      arrival_date_month %in% c("June", "July", "August") ~ "Summer",
      TRUE ~ "Fall"
    ),
    season = fct_relevel(season, "Winter", "Spring", "Summer", "Fall")
  )

hotels |>
  count(season, hotel, arrival_date_month) |>
  ggplot(aes(x = arrival_date_month, y = n, group = hotel, linetype = hotel)) +
  geom_line(linewidth = 0.8, color = "cornsilk4") +
  geom_point(aes(shape = season, color = season), size = 4, show.legend = FALSE) +
  scale_x_discrete(labels = month.abb) +
  scale_color_colorblind() +
  scale_shape_manual(values = c("circle", "square", "diamond", "triangle")) +
  labs(
    x = "Arrival month", y = "Number of bookings", linetype = NULL,
    title = "Number of monthly bookings",
    subtitle = "July 2015 to August 2017",
    caption = "Source: Antonio, Almeida and Nunes (2019) | TidyTuesday"
  ) +
  coord_cartesian(clip = "off") +
  theme(
    legend.position = c(0.12, 0.9),
    legend.box.background = element_rect(fill = "white", color = "white"),
    plot.subtitle = element_text(color = "cornsilk4"),
    plot.caption = element_text(color = "cornsilk4")
  )

A few takeaways

• forcats::fct_relevel() in a mutate() is useful for custom ordering of levels of a factor variable

• summarize() after group_by() with multiple variables results in a message about the grouping structure of the resulting data frame – the message can be supressed by defining .groups (e.g., .groups = "drop" or .groups = "keep")

• summarize() also lets you get away with being sloppy and not naming your new column, but that’s not recommended!

Rowwise operations

We want to calculate the total number of guests for each booking. Why does the following not work?

hotels |>
  select(adults, children, babies) |>
  mutate(guests = sum(c(adults, children, babies)))

# A tibble: 119,390 × 4
   adults children babies guests
    <dbl>    <dbl>  <dbl>  <dbl>
 1      2        0      0     NA
 2      2        0      0     NA
 3      1        0      0     NA
 4      1        0      0     NA
 5      2        0      0     NA
 6      2        0      0     NA
 7      2        0      0     NA
 8      2        0      0     NA
 9      2        0      0     NA
10      2        0      0     NA
# ℹ 119,380 more rows

Rowwise operations

hotels |>
  select(adults, children, babies) |>
  rowwise() |> 
  mutate(guests = sum(c(adults, children, babies))) |>
  filter(adults > 0, children > 0, babies > 0) # to show sum works

# A tibble: 172 × 4
# Rowwise: 
   adults children babies guests
    <dbl>    <dbl>  <dbl>  <dbl>
 1      2        1      1      4
 2      2        1      1      4
 3      2        1      1      4
 4      2        1      1      4
 5      2        1      1      4
 6      2        1      1      4
 7      2        1      1      4
 8      2        2      1      5
 9      2        2      1      5
10      1        2      1      4
# ℹ 162 more rows

Columnwise operations

Use across() combined with summarise() to calculate summary statistics for multiple columns at once:

hotels |>
  summarise(across(.cols = starts_with("stays"),  mean))

# A tibble: 1 × 2
  stays_in_weekend_nights stays_in_week_nights
                    <dbl>                <dbl>
1                   0.928                 2.50

hotels |>
  summarise(across(.cols = starts_with("stays"),  list(mean, sd)))

# A tibble: 1 × 4
  stays_in_weekend_nights_1 stays_in_weekend_nights_2
                      <dbl>                     <dbl>
1                     0.928                     0.999
# ℹ 2 more variables: stays_in_week_nights_1 <dbl>,
#   stays_in_week_nights_2 <dbl>

Select helpers

• starts_with(): Starts with a prefix
• ends_with(): Ends with a suffix
• contains(): Contains a literal string
• num_range(): Matches a numerical range like x01, x02, x03
• one_of(): Matches variable names in a character vector
• everything(): Matches all variables
• last_col(): Select last variable, possibly with an offset
• matches(): Matches a regular expression (a sequence of symbols/characters expressing a string/pattern to be searched for within text)

See help for any of these functions for more info, e.g. ?everything.

Columnwise operations

hotels |>
  group_by(hotel, is_canceled) |>
  summarise(
    across(.cols = starts_with("stays"),  list(mean = mean, sd = sd), .names = "{.fn}_{.col}") 
    )

# A tibble: 4 × 6
# Groups:   hotel [2]
  hotel is_canceled mean_stays_in_weeken…¹ sd_stays_in_weekend_…²
  <chr>       <dbl>                  <dbl>                  <dbl>
1 City…           0                  0.801                  0.862
2 City…           1                  0.788                  0.917
3 Reso…           0                  1.13                   1.14 
4 Reso…           1                  1.34                   1.14 
# ℹ abbreviated names: ¹​mean_stays_in_weekend_nights,
#   ²​sd_stays_in_weekend_nights
# ℹ 2 more variables: mean_stays_in_week_nights <dbl>,
#   sd_stays_in_week_nights <dbl>

Columnwise operations

hotels |>
  group_by(hotel, is_canceled) |>
  summarise(
    across(.cols = starts_with("stays"),  list(mean = mean, sd = sd), .names = "{.fn}_{.col}"),
    .groups = "drop"
    )

# A tibble: 4 × 6
  hotel is_canceled mean_stays_in_weeken…¹ sd_stays_in_weekend_…²
  <chr>       <dbl>                  <dbl>                  <dbl>
1 City…           0                  0.801                  0.862
2 City…           1                  0.788                  0.917
3 Reso…           0                  1.13                   1.14 
4 Reso…           1                  1.34                   1.14 
# ℹ abbreviated names: ¹​mean_stays_in_weekend_nights,
#   ²​sd_stays_in_weekend_nights
# ℹ 2 more variables: mean_stays_in_week_nights <dbl>,
#   sd_stays_in_week_nights <dbl>

Setup for next example: hotel_summary

hotels_summary <- hotels |>
  group_by(hotel, is_canceled) |>
  summarise(
    across(
      .cols = starts_with("stays"),
      list(mean = mean),
      .names = "{.fn}_{.col}"
    ),
    .groups = "drop"
  )

hotels_summary

# A tibble: 4 × 4
  hotel is_canceled mean_stays_in_weeken…¹ mean_stays_in_week_n…²
  <chr>       <dbl>                  <dbl>                  <dbl>
1 City…           0                  0.801                   2.12
2 City…           1                  0.788                   2.27
3 Reso…           0                  1.13                    3.01
4 Reso…           1                  1.34                    3.44
# ℹ abbreviated names: ¹​mean_stays_in_weekend_nights,
#   ²​mean_stays_in_week_nights

Which variables are plotted in the following visualization? Which aesthetics are they mapped to? Recreate the visualization.

Livecoding

Reveal below for code developed during live coding session.

hotels_summary |>
  mutate(is_canceled = if_else(is_canceled == 0, "Not canceled", "Canceled")) |>
  pivot_longer(cols = starts_with("mean"),
               names_to = "day_type",
               values_to = "mean_stays",
               names_prefix = "mean_stays_in_") |>
  mutate(
    day_type = if_else(str_detect(day_type, "weekend"), "Weekend", "Weekday")
    ) |>
  ggplot(aes(x = str_wrap(is_canceled, 10), y = mean_stays, 
             group = hotel, color = hotel)) +
  geom_point(show.legend = FALSE) +
  geom_line(aes(linetype = hotel), linewidth = 1) +
  facet_wrap(~day_type) +
  labs(
    x = "Booking status",
    y = "Mean number of\nnights of stay",
    color = NULL, linetype = NULL,
    title = "Mean number of stays",
    subtitle = "By hotel type and booking status",
    caption = "Source: Antonio, Almeida and Nunes (2019) | TidyTuesday"
  ) +
  scale_color_manual(values = c("cornsilk4", "deepskyblue3")) +
  scale_y_continuous(limits = c(0, 4), breaks = 0:4) +
  theme(legend.position = "bottom")

pivot_wider() and pivot_longer()

• From tidyr
• Incredibly useful for reshaping for plotting
• Lots of extra arguments to help with reshaping pain!
• Refer to pivoting vignette when needed

Stats

Stats < > geoms

• Statistical transformation (stat) transforms the data, typically by summarizing
• Many of ggplot2’s stats are used behind the scenes to generate many important geoms

stat	geom
stat_bin()	geom_bar(), geom_freqpoly(), geom_histogram()
stat_bin2d()	geom_bin2d()
stat_bindot()	geom_dotplot()
stat_binhex()	geom_hex()
stat_boxplot()	geom_boxplot()
stat_contour()	geom_contour()
stat_quantile()	geom_quantile()
stat_smooth()	geom_smooth()
stat_sum()	geom_count()

Layering with stats

hotels |>
  filter(adr < 4000) |>
  ggplot(aes(x = arrival_date_month, y = adr)) +
  geom_point(alpha = 0.5) +
  stat_summary(
    geom = "point", fun = "median", 
    colour = "red", size = 5, pch = 4, stroke = 2
  ) +
  facet_wrap(~ hotel, ncol = 1)

Alternate: layering with stats

hotels |>
  filter(adr < 4000) |>
  ggplot(aes(x = arrival_date_month, y = adr)) +
  geom_point(alpha = 0.5) +
  geom_point(
    stat = "summary", fun = "median", 
    colour = "red", size = 5, pch = 4, stroke = 2
  ) +
  facet_wrap(~ hotel, ncol = 1)

Statistical transformations

What can you say about the distribution of price from the following QQ plot?

hotels |>
  filter(adr < 4000) |>
  ggplot(aes(sample = adr)) +
  stat_qq() +
  stat_qq_line() +
  labs(y = "adr")