Combining Data Manipulations in dplyr

Introduction

• Combine a series of data manipulation actions
• Do each action in sequential order

Intermediate variables

• Run a command
• Store the output in a variable
• Use that variable later in the code

• Repeat

• Obtain the data for only DS, sorted by year, with only the year and and weight columns

ds_data <- filter(surveys, species_id == "DS", !is.na(weight))
ds_data_by_year <- arrange(ds_data, year)
ds_weight_by_year <- select(ds_data_by_year, year, weight)

Do Portal Data Manipulation Exercise

Pipes

• Intermediate variables can get cumbersome if their are lots of steps.
• |> or %>% (“pipe”) takes the output of one command and passes it as input to the next command
• Want to take the mean of a vector
• Normally we would run the mean function with the vector as the input:

x = c(1, 2, 3)
mean(x)

• Instead we could pipe the vector into the function

x |> mean()
x %>% mean

• So x becomes the first argument in mean
• If we want to add other arguments they get added to the function call

x = c(1, 2, 3, NA)
mean(x, na.rm = TRUE)
x |> mean(na.rm = TRUE)

• Questions?

surveys |>
  filter(species_id == "DS", !is.na(weight))

ds_weight_by_year <- surveys |>
  filter(species_id == "DS", !is.na(weight)) |>
  arrange(year) |>
  select(year, weight)

Do Portal Data Manipulation Pipes.

The magrittr pipe

• You will also see another type of pipe character %>%
• This is the original pipe in R and you had to load the magrittr package to use it (this gets loaded automatically by dplyr)
• Either pipe is fine for this class
  • |> will work everywhere as long as you have a new enough version of R
  • magrittr has some fancier functionality that may be useful in some cases

Keyboard Shortcut

• Shortcut: Ctrl-Shift-m
• You can change this to give the base R pipe
  • Tools -> Global Options -> Code -> Use native pipe operator

What if I want to pipe to an argument other than the first argument

surveys |>
  filter(species_id == "DS", !is.na(weight)) |>
  arrange(year) |>
  select(year, weight) |>
  lm(weight ~ year, data = .)