Multi-Page dplyr

Andy Grogan-Kaylor
2021-10-02

Background

dplyr is a very powerful R library for managing and processing data.1

While dplyr is very powerful, learning to use dplyr can be very confusing. This guide aims to present some of the most common dplyr functions and commands in the form of a brief cheatsheet.

library(dplyr)

Simulated Data

year x y z
2019 NA Group B 90
2020 50 Group C 90
2020 NA Group A 100
2021 70 Group A 110
2021 80 Group B 120

Piping

Pipes %>% connect pieces of a command e.g. data to data wrangling to a graph command.

dplyr commands will often look something like the outline below.

mydata %>% 
  data_wrangling %>% 
  more_data_wrangling %>%
  graph_command

Aggregate Data: group_by() & summarise()

Notice how when aggregating data, we need to be explicit about whether we are removing NA values.

mynewdata <- mydata %>% 
  group_by(year) %>% # group by y
  summarise(mean_x = mean(x, na.rm = TRUE), # mean of x; removing NA's
            n = n()) # count up
year mean_x n
2019 NA 1
2020 50 2
2021 75 2

Select A Subset of Variables: select()

mynewdata <- mydata %>% 
  select(x,y) # select only x and y
x y
NA Group B
50 Group C
NA Group A
70 Group A
80 Group B

Filter A Subset of Rows: filter()

mynewdata <- mydata %>% 
  filter(year > 2020) # filter on year
year x y z
2021 70 Group A 110
2021 80 Group B 120

Create New Variables: mutate()

mynewdata <- mydata %>% 
  mutate(myscale = x + z) # create a new variable e.g. a scale
year x y z myscale
2019 NA Group B 90 NA
2020 50 Group C 90 140
2020 NA Group A 100 NA
2021 70 Group A 110 180
2021 80 Group B 120 200

Recode Variables: mutate()

Continuous Into Categorical: mutate() & cut()

mynewdata <- mydata %>% 
  mutate(zcategorical = cut(z, # cut at breaks
                            breaks=c(-Inf, 100, Inf), 
               labels = c("low", "high")))
year x y z zcategorical
2019 NA Group B 90 low
2020 50 Group C 90 low
2020 NA Group A 100 low
2021 70 Group A 110 high
2021 80 Group B 120 high

Categorical Into Categorical: mutate() & recode()

mynewdata <- mydata %>% 
  mutate(yrecoded = dplyr::recode(y, # recode values
                         "Group A" = "Red Group",
                         "Group B" = "Blue Group", 
                         .default = "Other"))
year x y z yrecoded
2019 NA Group B 90 Blue Group
2020 50 Group C 90 Other
2020 NA Group A 100 Red Group
2021 70 Group A 110 Red Group
2021 80 Group B 120 Blue Group

Rename Variables: rename()

newdata <- mydata %>% 
  rename(age = x, # rename
         mental_health = z)
year age y mental_health
2019 NA Group B 90
2020 50 Group C 90
2020 NA Group A 100
2021 70 Group A 110
2021 80 Group B 120

Drop Missing Values: filter()

newdata <- mydata %>% 
  filter(!is.na(x)) # filter by x is not missing
year x y z
2020 50 Group C 90
2021 70 Group A 110
2021 80 Group B 120

Random Sample

newdata <- mydata %>% 
  sample_frac(.5) # fraction of data to sample
year x y z
2020 50 Group C 90
2020 NA Group A 100

Connecting To Other Packages Like ggplot

Notice how, in the code below, I never actually create the new data set mynewdata. I simply pipe mydata into a dplyr command, and pipe the result directly to ggplot2.

library(ggplot2)

mydata %>% # my data
  mutate(myscale = x + z) %>% # dplyr command to make new variable 
  filter(y != "Group C") %>% # filter on values of y
  ggplot(aes(x = year, # the rest is ggplot
             fill = y)) +
  geom_bar() +
  scale_fill_viridis_d(option = "cividis") + # beautiful cividis colors
  theme_minimal() + # minimal theme
  labs(title = "Group by Year") + # labels
  theme(axis.text.x = element_text(size = 10, # tweak theme
                                   angle = 90))

  1. The origins of the name dplyr seem somewhat obscure, but I sometimes think of this package as the data plyers.↩︎