What is tidy data?

Reshaping Data with tidyr

Jeroen Boeye

Head of Machine Learning, Faktion

Happy families are all alike, but every unhappy family is unhappy in its own way.

Leo Tolstoy

Tidy datasets are all alike, but every messy dataset is messy in its own way.

Hadley Wickham

Rectangular data

Structure

Columns
Rows
Cells

Tidy sample

Tidy data, variables

Structure

Columns hold variables
Rows
Cells

Tidy sample variables

Tidy data, observations

Structure

Columns hold variables
Rows hold observations
Cells

Tidy sample observations

Tidy data, values

Structure

Columns hold variables
Rows hold observations
Cells hold values

Tidy sample values

dplyr recap

character_df

# A tibble: 4 x 3
  name           homeworld species
  <chr>          <chr>     <chr>  
1 Luke Skywalker Tatooine  Human  
2 R2-D2          Naboo     Droid  
3 Darth Vader    Tatooine  Human  
4 Obi-Wan Kenobi Stewjon   Human

dplyr recap: select()

character_df %>% 
  select(name, homeworld)

# A tibble: 4 x 2
  name           homeworld
  <chr>          <chr>    
1 Luke Skywalker Tatooine 
2 R2-D2          Naboo    
3 Darth Vader    Tatooine 
4 Obi-Wan Kenobi Stewjon

dplyr recap: filter()

character_df %>% 
  filter(homeworld == "Tatooine")

# A tibble: 2 x 3
  name           homeworld species
  <chr>          <chr>     <chr>  
1 Luke Skywalker Tatooine  Human  
2 Darth Vader    Tatooine  Human

dplyr recap: mutate()

character_df %>% 
  mutate(is_human = species == "Human")

# A tibble: 4 x 4
  name           homeworld species is_human
  <chr>          <chr>     <chr>   <lgl>   
1 Luke Skywalker Tatooine  Human   TRUE    
2 R2-D2          Naboo     Droid   FALSE   
3 Darth Vader    Tatooine  Human   TRUE    
4 Obi-Wan Kenobi Stewjon   Human   TRUE

dplyr recap: group_by() and summarize()

character_df %>% 
  group_by(homeworld) %>% 
  summarize(n = n())

# A tibble: 3 x 2
  homeworld     n
  <chr>     <int>
1 Naboo         1
2 Stewjon       1
3 Tatooine      2

magrittr logo

¹ magrittr.tidyverse.org

dplyr logo

tidyr logo

¹ www.tidyverse.org

Multiple variables in a single column

population_df

# A tibble: 4 x 2
  country                 population
  <chr>                        <dbl>
1 Brazil, South America        210. 
2 Nepal, Asia                   28.1
3 Senegal, Africa               15.8
4 Australia, Oceania            25.0

Separating variables over two columns

population_df %>% 
  separate(country, into = c("country", "continent"), sep = ", ")

# A tibble: 4 x 3
  country   continent      population
  <chr>     <chr>               <dbl>
1 Brazil    South America       210. 
2 Nepal     Asia                 28.1
3 Senegal   Africa               15.8
4 Australia Oceania              25.0

Let's practice!

Reshaping Data with tidyr