UQRUG 69

Overview: ggplot2 and dplyr news
Questions: recoding, data structures, missing data, shell

In March, 13 ruggers met to learn about what’s new in the latest ggplot2 and dplyr major releases.

Topic: What’s new in ggplot2 and dplyr?

What are some of the main new features in ggplot2 4.0 and dplyr 1.2?

Recording of the presentation

Script written during the presentation:

#### ggplot2 4.0 ####

library(ggplot2)

#### themes ####

# originally:
ggplot(penguins, aes(x = bill_len, y = flipper_len)) +
  geom_point(colour = "brown") +
  geom_density2d(colour = "orange") +
  geom_smooth(colour = "orange") +
  theme_minimal() +
  theme(plot.background = element_rect(fill = "beige"))

# now, you can use "paper", "ink" and "accent":
ggplot(penguins, aes(x = bill_len, y = flipper_len)) +
  geom_point() +
  geom_density2d() +
  geom_smooth() +
  theme_minimal(paper = "beige",
                ink = "brown",
                accent = "orange")

# define geometry options into the theme
ggplot(penguins, aes(x = species, y = body_mass)) +
  geom_jitter() +
  geom_boxplot() +
  theme(geom = element_geom(ink = "pink",
                            paper = "purple"))

# aside: styling changes in geometries
ggplot(penguins, aes(x = species, y = body_mass)) +
  geom_jitter() +
  geom_boxplot(box.linewidth = 3,
               whisker.colour = "black") +
  theme(geom = element_geom(ink = "pink",
                            paper = "purple"))

# other geoms have styling updates. We always could apply gradient to points:
ggplot(economics, aes(date, unemploy, colour = uempmed)) +
  geom_point()
# now you can also for an area geom:
ggplot(economics, aes(date, unemploy, fill = uempmed)) +
  geom_area()
# other styling updates: violin, ribbon, label

# back to themes
# distinction between borders and lines:
ggplot(penguins, aes(x = bill_len, y = flipper_len)) +
  geom_bin2d(bins = 10, colour = "black") +
  geom_smooth() +
  # change the direction of filling with facets
  facet_wrap(vars(species), dir = "br") +
  theme(geom = element_geom(borderwidth = 0.5,
                            linewidth = 2),
        # change the palette in the theme:
        palette.colour.continuous = c("pink", "purple")) +
  # fixed size for panels:
  theme_sub_panel(widths = unit(1:3, "cm"))
# different to the size of image in ggsave()!

#### labels ####

# column labels:
attr(penguins$flipper_len, "label") <- "Flipper length (cm)"
# see it
penguins$flipper_len

# now used automatically for titles on the plot:
ggplot(penguins, aes(y = bill_len, x = flipper_len)) +
  geom_bin2d(bins = 10, colour = "black") +
  geom_smooth()
# labs(dictionary = ...) also works if they are stored in a separate table

#### scale breaks ####
ggplot(penguins, aes(y = bill_len, x = flipper_len)) +
  geom_bin2d(bins = 10, colour = "black") +
  geom_smooth() +
  scale_x_continuous(breaks = c("two hundred" = 200,
                                "thirty more" = 230))

#### aesthetics in positions of geometry ####

ggplot(penguins, aes(x = species, fill = island)) +
  geom_bar(position = position_dodge2(preserve = "single"))

#### dplyr ####

library(dplyr)

#### filter_out() ####
penguins |> 
  as_tibble() |> 
  filter(species != "Adelie")
# now, we can use filter_out():
penguins |> 
  as_tibble() |> 
  filter_out(species == "Adelie")

# filter() is for keeping rows, it drops NAs
# filter_out() is for removing rows, it KEEPS NAs.

#### recoding / replacing ####

# in addition to case_when(), we now have more functions to recode and replace values.

# recode: create a new column based on an existing one
# replace: replace values in existing column

# original:
recoded <- penguins |> 
  mutate(spec_short = case_when(species == "Adelie" ~ "Ad",
                                species == "Gentoo" ~ "Gen",
                                species == "Chinstrap" ~ "Chin"))
# with recode_values() instead:
recoded <- penguins |> 
  mutate(spec_short = recode_values(species,
                                    "Adelie" ~ "Ad",
                                    "Gentoo" ~ "Gen",
                                    "Chinstrap" ~ "Chin"))

Here are the corresponding blog posts to learn more about these releases:

• ggplot2
• dplyr

Questions

We also discussed recoding variable a bit further, as well as R data structure / types, dealing with missing data, and executing shell commands from R:

# recode body mass into categorical variable
range(penguins$body_mass, na.rm = TRUE)

# from continuous to discrete
categorised <- penguins |> 
  mutate(mass_cat = case_when(body_mass < 3000 ~ "light",
                              body_mass < 5000 ~ "mid",
                              is.na(body_mass) ~ NA,
                              .default = "heavy"),
         .after = body_mass)

# dealing with NAs
library(tidyr)
replace_na(penguins, list(body_mass = 0,
                          bill_dep = mean(penguins$bill_dep,
                                          na.rm = TRUE)))

# vectors:
single_value <- 1
my_vector <- c(1, 5, 3, NA, 1)
typeof(my_vector)
my_vector[5] <- 1.45
# convert to a different type:
my_int_vector <- as.integer(my_vector)
my_int_vector[5] <- 1.45 # silently converts to double

# matrices:
matrix(1:10, 2)

# list: more flexible
my_list <- list(my_vector,
                penguins,
                another_number = 1,
                list(1, 2, 5, "a string"))

# technically, a dataframe is a list
# (of vectors of same length)

# run bash/shell script from R:
system("echo 'hi'")