Creating custom Quantiles within data frame?

Question

If i have the following table:

tibble(year = c("2020", "2020", "2020","2021", "2021", "2021"),
       website  = c("facebook", "google", "youtube","facebook", "google", "youtube"), 
       method = c("laptop", "laptop", "laptop", "mobile", "mobile", "mobile"), 
       values = c(10,30,60, 90,25, 40))

How can i try to go about creating a column based on a custom q-tile of the numbers in the value column.

For example, if i have the following custom q-tile conditions:

Risky - > 50% Neither - 25-50% Safe - <25%

These are basically saying for the numbers in the values column, calculate what their ranking based on the q-tile conditions above and give them a rank value of 1,2,3 accordingly.

The final table should look like this:

tibble(year = c("2020", "2020", "2020","2021", "2021", "2021"),
       website  = c("facebook", "google", "youtube","facebook", "google", "youtube"), 
       method = c("laptop", "laptop", "laptop", "mobile", "mobile", "mobile"), 
       values = c(10,30,60, 90,25, 40), 
       rank = c(3,2,1,1,3,2))

I know that the table will have to be grouped by year and method along, so the code will look like this:

df %>% group_by(year, method) %>% mutate(rank = quantile(???))

Answer 1

You can use quantile(x, c(0.25, 0.5)) to get cutpoints and pass them into findInterval(). Note that findInterval() is similar to cut(*, labels = FALSE) but is more efficient.

library(dplyr)

df %>%
  group_by(year, method) %>%
  mutate(rank = findInterval(-values, quantile(-values, c(0.25, 0.5)), left.open = TRUE) + 1) %>%
  ungroup()

# # A tibble: 6 × 5
#   year  website  method values  rank
#   <chr> <chr>    <chr>   <dbl> <dbl>
# 1 2020  facebook laptop     10     3
# 2 2020  google   laptop     30     2
# 3 2020  youtube  laptop     60     1
# 4 2021  facebook mobile     90     1
# 5 2021  google   mobile     25     3
# 6 2021  youtube  mobile     40     2

---

If you want labels rather than ranks, use cut():

df %>%
  group_by(year, method) %>%
  mutate(rank = cut(values, quantile(values, c(0, 0.25, 0.5, 1)),
                    c("Safe", "Neither", "Risky"), include.lowest = TRUE)) %>%
  ungroup()

# # A tibble: 6 × 5
#   year  website  method values rank   
#   <chr> <chr>    <chr>   <dbl> <fct>  
# 1 2020  facebook laptop     10 Safe   
# 2 2020  google   laptop     30 Neither
# 3 2020  youtube  laptop     60 Risky  
# 4 2021  facebook mobile     90 Risky  
# 5 2021  google   mobile     25 Safe   
# 6 2021  youtube  mobile     40 Neither

Answer 2

You can use the ntile function from dplyr to create quantiles:

library(dplyr)
df %>%
  group_by(year, method) %>%
  mutate(rank = ntile(values, 4))

Output:

# A tibble: 6 × 5
# Groups:   year, method [2]
  year  website  method values  rank
  <chr> <chr>    <chr>   <dbl> <int>
1 2020  facebook laptop     10     1
2 2020  google   laptop     30     2
3 2020  youtube  laptop     60     3
4 2021  facebook mobile     90     3
5 2021  google   mobile     25     1
6 2021  youtube  mobile     40     2

Answer 3

df %>%
  group_by(year, method) %>%
  mutate(rank = rank(-cut(values, breaks = c(-Inf, quantile(values, probs = c(0.25, 0.50), names = F), Inf), labels = F)))

# # A tibble: 6 x 5
# # Groups:   year, method [2]
#   year  website  method values  rank
#   <chr> <chr>    <chr>   <dbl> <dbl>
# 1 2020  facebook laptop     10     3
# 2 2020  google   laptop     30     2
# 3 2020  youtube  laptop     60     1
# 4 2021  facebook mobile     90     1
# 5 2021  google   mobile     25     3
# 6 2021  youtube  mobile     40     2

Answer 4

A {santoku} one-liner:

mutate(df,
  rank = santoku::chop_quantiles(rank, c(0.25, 0.5), 
                                   labels = c("Safe", "Neither", "Risky"))
)