Creating unique combinations of vectors

Reshaping Data with tidyr

Jeroen Boeye

Head of Machine Learning, Faktion

The early atomic era: 1945 - 1954

nuke_df

# A tibble: 13 x 3
   country             year n_bombs
   <chr>              <int>   <int>
 1 United States       1945       3
 2 United States       1946       2
 3 United States       1948       3
 4 Russian Federation  1949       1
 5 Russian Federation  1951       2
 6 United States       1951      16
# ... with 7 more rows

The expand_grid() function

full_df <- expand_grid(
  year = 1945:1954, 
  country = c(
    "Russian Federation", 
    "United Kingdom", 
    "United States")
  )

full_df

# A tibble: 30 x 2
    year country           
   <int> <chr>             
 1  1945 Russian Federation
 2  1945 United Kingdom    
 3  1945 United States     
 4  1946 Russian Federation
 5  1946 United Kingdom    
 6  1946 United States     
 7  1947 Russian Federation
 8  1947 United Kingdom    
# ... with 22 more rows

right_join() with a tibble of unique combinations

nuke_df %>% 
  right_join(
    full_df, 
    by = c("country", "year")
  ) %>%
  arrange(year)

# A tibble: 30 x 3
   country             year n_bombs
   <chr>              <int>   <int>
 1 United States       1945       3
 2 Russian Federation  1945      NA
 3 United Kingdom      1945      NA
 4 United States       1946       2
 5 Russian Federation  1946      NA
 6 United Kingdom      1946      NA
 7 Russian Federation  1947      NA
 8 United Kingdom      1947      NA
# ... with 22 more rows

right_join() with a tibble of unique combinations

nuke_df %>% 
  right_join(
    full_df, 
    by = c("country", "year")
  ) %>%
  arrange(year) %>% 
  replace_na(list(n_bombs = 0L))

# A tibble: 30 x 3
   country             year n_bombs
   <chr>              <int>   <int>
 1 United States       1945       3
 2 Russian Federation  1945       0
 3 United Kingdom      1945       0
 4 United States       1946       2
 5 Russian Federation  1946       0
 6 United Kingdom      1946       0
 7 Russian Federation  1947       0
 8 United Kingdom      1947       0
# ... with 22 more rows

anti_join() to select missing observations

full_df %>% 
  anti_join(
    nuke_df, 
    by = c("country", "year")
  )

# A tibble: 17 x 2
    year country           
   <int> <chr>             
 1  1945 Russian Federation
 2  1945 United Kingdom    
 3  1946 Russian Federation
 4  1946 United Kingdom    
 5  1947 Russian Federation
 6  1947 United Kingdom    
 7  1947 United States     
 8  1948 Russian Federation
# ... with 9 more rows

Let's practice!

Reshaping Data with tidyr