在数据框中创建自定义分位数？

Question

如果我有以下 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))

我如何尝试根据值列中数字的自定义 q-tile 创建列。

例如，如果我有以下自定义 q-tile 条件：

风险 - > 50% 两者都不是 - 25-50% 安全 - <25%

这些基本上是说对于值列中的数字，根据上面的 q-tile 条件计算他们的排名，并相应地给他们一个排名值 1,2,3。

最后的 table 应该是这样的：

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))

我知道 table 必须按年份和方法分组，因此代码如下所示：

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

Answer 1

您可以使用 dplyr 中的 ntile 函数创建分位数：

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

输出：

# 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 2

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 3

您可以使用 quantile(x, c(0.25, 0.5)) 获取切点并将它们传递给 findInterval()。请注意 findInterval() 类似于 cut(*, labels = FALSE) 但效率更高。

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

如果您想要标签而不是排名，请使用 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 4

A {santoku} one-liner:

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

在数据框中创建自定义分位数？

Creating custom Quantiles within data frame?

r

dataframe

dplyr

tidyverse