TidyTuesday: Cocktails

Standardize cases

bc_raw %>% 
  count(ingredient, sort = TRUE) %>% 
  filter(str_detect(ingredient, "red pepper sauce"))

## # A tibble: 2 x 2
##   ingredient               n
##   <chr>                <int>
## 1 Hot red pepper sauce     4
## 2 hot red pepper sauce     1

Let’s fix that by making all ingredient values to lower case:

bc <- bc_raw %>% 
  mutate(ingredient = str_to_lower(ingredient)) %>% 
  distinct() %>% 
  select(name, ingredient)

And to make sure it works…

bc %>% 
  count(ingredient, sort = TRUE) %>% 
  filter(str_detect(ingredient, "red pepper sauce"))

## # A tibble: 1 x 2
##   ingredient               n
##   <chr>                <int>
## 1 hot red pepper sauce     5

Fix untidy data

To follow “tidy” principles, we need one row per observation.

bc %>% 
  filter(str_detect(ingredient, ",")) 

## # A tibble: 85 x 2
##    name                  ingredient                                             
##    <chr>                 <chr>                                                  
##  1 John Collins          orange and lemon wheels, maraschino cherry             
##  2 Irish Shillelagh      fresh raspberries and strawberries, 2 peach slices, ma…
##  3 Underneath The Mango… lime wedge, sweet chili powder                         
##  4 Emperor Norton's Mis… fresh strawberries, cut in halves                      
##  5 Toasted Drop          lemon wedge, cinnamon sugar                            
##  6 Stockholm 75          lemon wedge, superfine sugar                           
##  7 Salty Dog             lemon wedge, coarse salt                               
##  8 Rouxby Red            for glass lemon wedge, coarse salt                     
##  9 Redhead Martini       strawberries, cut into halves                          
## 10 Canadian Breeze       pineapple wedge, maraschino cherry                     
## # … with 75 more rows

We can reformat this by separating the commas then adding a new row for each cocktail per ingredient.

# CLEAN DATAFRAME
bc_tidy <- bc %>% 
  filter(!str_detect(ingredient, ","))

# EXTRACT UNTIDY DATA THEN CLEAN
bc_untidy <- bc %>% 
  filter(str_detect(ingredient, ",")) %>% 
  mutate(ingredient = str_split(ingredient, ", ")) %>% 
  unnest(ingredient)

# COMBINE BOTH DATAFRAMES
bc_clean <- rbind(bc_tidy, bc_untidy) %>% 
  distinct()

bc_untidy

## # A tibble: 193 x 2
##    name                      ingredient                        
##    <chr>                     <chr>                             
##  1 John Collins              orange and lemon wheels           
##  2 John Collins              maraschino cherry                 
##  3 Irish Shillelagh          fresh raspberries and strawberries
##  4 Irish Shillelagh          2 peach slices                    
##  5 Irish Shillelagh          maraschino cherry                 
##  6 Underneath The Mango Tree lime wedge                        
##  7 Underneath The Mango Tree sweet chili powder                
##  8 Emperor Norton's Mistress fresh strawberries                
##  9 Emperor Norton's Mistress cut in halves                     
## 10 Toasted Drop              lemon wedge                       
## # … with 183 more rows

Reduce cardinality

Our dataset has more than 550 unique ingredients so let’s trim that down to the ingredients that are used in ten or more cocktails.

bc_clean %>% 
  distinct(ingredient) %>% 
  count()

## # A tibble: 1 x 1
##       n
##   <int>
## 1   553

n_ingredients <- bc_clean %>% 
  count(ingredient, sort = TRUE) %>% 
  filter(n > 10)

Normalize ingredients

Similar to our case statement section above, let’s make sure our ingredients are consolidated to the same format.

df <- bc_clean %>% 
  inner_join(n_ingredients) %>% 
  select(-n) %>% 
  mutate(ingredient = str_replace_all(ingredient, "-", "_"),
         ingredient = str_replace_all(ingredient, " ", "_"),
         ingredient = str_replace_all(ingredient, "old_mr._boston_", ""),
         ingredient = str_replace_all(ingredient, "old_thompson_", "")) 

Format data for MCA

df_mca_processed <- df %>% 
  mutate(value = 1) %>%
  pivot_wider(names_from = ingredient) %>% 
  replace(is.na(.), 0) %>% 
  select(-name) %>%
  mutate_if(is.double, as.factor)

mca_results <- MCA(df_mca_processed, graph = FALSE)

Shape data to tidy structure

mca_df <- data.frame(mca_results$var$coord)

mca_final <- rownames_to_column(mca_df, var = "rowname") %>% 
  as_tibble() %>% 
  filter(str_detect(rowname, "_1")) %>% 
  mutate(variable = str_replace_all(rowname, "_1", "")) %>% 
  select(variable, everything(), -rowname) %>% 
  mutate(highlight = case_when(str_detect(variable, "gin") ~ "gin",
                               str_detect(variable, "rum") ~ "rum",
                               str_detect(variable, "vodka") ~ "vodka",
                               str_detect(variable, "whiskey") ~ "whiskey",
                               str_detect(variable, "brandy") ~ "brandy",
                               str_detect(variable, "bourbon") ~ "bourbon",
                               str_detect(variable, "tequila") ~ "tequila"))

Final plot

mca_final %>% 
  ggplot(aes(x = Dim.1, y = Dim.2, label = variable, color = highlight)) +
  geom_density2d(color = "gray90") +
  geom_point(show.legend = FALSE) +
  geom_text_repel(show.legend = FALSE) +
  labs(x = "D1", y = "D2", 
       title = "Multiple correspondence analysis (MCA) on the most common cocktail ingredients",
       subtitle = "Closer points suggest they are typically mixed together",
       caption  = "by: @eeysirhc\nsource: Mr. Boston Bartender's Guide") +
  theme_minimal(base_size = 15) +
  theme(axis.text.y = element_blank(),
        axis.text.x = element_blank(),
        panel.grid.major = element_blank(),
        panel.grid.minor = element_blank())