Tidy Data with tidyr

Tidy data follows three rules:

1. Each variable is a column.
2. Each observation is a row.
3. Each value is a cell.

Most tidyverse tools expect tidy data. Real data often isn't: a common example is one column per year. tidyr reshapes it.

pivot_longer — wide to long

library(tidyr)

sales <- tibble::tibble(
  region = c("North", "South"),
  `2023` = c(120, 95),
  `2024` = c(135, 110)
)
sales

sales_long <- sales |>
  pivot_longer(cols = c(`2023`, `2024`), names_to = "year", values_to = "sales")
sales_long

# A tibble: 2 × 3
  region `2023` `2024`
  <chr>   <dbl>  <dbl>
1 North     120    135
2 South      95    110
# A tibble: 4 × 3
  region year  sales
  <chr>  <chr> <dbl>
1 North  2023    120
2 North  2024    135
3 South  2023     95
4 South  2024    110

"Year" is a variable, so it should be one column — now it is. The long form is what ggplot2 and dplyr's group_by work best with. (Column names that start with a number need backticks.)

pivot_wider — long to wide

The reverse, useful for presenting a summary table:

library(tidyr)

long <- tibble::tibble(
  region = c("North", "North", "South", "South"),
  year = c("2023", "2024", "2023", "2024"),
  sales = c(120, 135, 95, 110)
)

long |> pivot_wider(names_from = year, values_from = sales)

# A tibble: 2 × 3
  region `2023` `2024`
  <chr>   <dbl>  <dbl>
1 North     120    135
2 South      95    110

Splitting and handling missing values

library(tidyr)

people <- tibble::tibble(full_name = c("Ada Lovelace", "Alan Turing"), score = c(90, NA))

people |> separate_wider_delim(full_name, delim = " ", names = c("first", "last"))
people |> replace_na(list(score = 0))
people |> drop_na(score)

# A tibble: 2 × 3
  first last     score
  <chr> <chr>    <dbl>
1 Ada   Lovelace    90
2 Alan  Turing      NA
# A tibble: 2 × 2
  full_name    score
  <chr>        <dbl>
1 Ada Lovelace    90
2 Alan Turing      0
# A tibble: 1 × 2
  full_name    score
  <chr>        <dbl>
1 Ada Lovelace    90

Common mistakes

• Keeping data wide and then struggling to plot or summarise it. If a piece of information (a year, a category) is stored in column names, pivot it longer first.