我有一个按组嵌套的数据框。我想将变量“ x”从其原始值转换为分位数位置(20%,40%,60%,80%,100%或1、2、3、4、5)。
这里是我正在使用的数据的示例:
df <- data.frame(x=c(1, 5, 21, 24, 43, 47, 56, 59, 68, 69, 11, 15, 25, 27, 48, 49, 51, 55, 61, 67),
y=c("A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "B", "B", "B", "B", "B", "B", "B", "B", "B", "B"))
这是我尝试过的:
df$z <- aggregate(df$x, by = list(df$y), FUN = function(x) quantile(x, probs = c(0.2, 0.4, 0.6, 0.8, 1), na.rm = T))
本质上,我想创建一个看起来像这样的新变量:
df$z <- c(1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5)
在分组的数据框架上,您可以使用dplyr::ntile()
:
library(dplyr)
df %>%
group_by(y) %>%
mutate(z = ntile(x, 5))
# A tibble: 20 x 3
# Groups: y [2]
x y z
<dbl> <fct> <int>
1 1 A 1
2 5 A 1
3 21 A 2
4 24 A 2
5 43 A 3
6 47 A 3
7 56 A 4
8 59 A 4
9 68 A 5
10 69 A 5
11 11 B 1
12 15 B 1
13 25 B 2
14 27 B 2
15 48 B 3
16 49 B 3
17 51 B 4
18 55 B 4
19 61 B 5
20 67 B 5
我们可以将cut
和breaks
用作quantile
library(dplyr)
df %>%
group_by(y) %>%
mutate(z = as.integer(cut(x, breaks = c(-Inf,
quantile(x, probs = c(0.2, 0.4, 0.6, 0.8, 1), na.rm = TRUE)))))
# A tibble: 20 x 3
# Groups: y [2]
# x y z
# <dbl> <fct> <int>
# 1 1 A 1
# 2 5 A 1
# 3 21 A 2
# 4 24 A 2
# 5 43 A 3
# 6 47 A 3
# 7 56 A 4
# 8 59 A 4
# 9 68 A 5
#10 69 A 5
#11 11 B 1
#12 15 B 1
#13 25 B 2
#14 27 B 2
#15 48 B 3
#16 49 B 3
#17 51 B 4
#18 55 B 4
#19 61 B 5
#20 67 B 5
或与base R
一起使用ave
with(df, ave(x, y, FUN = function(u) as.integer(cut(u, breaks = c(-Inf,
quantile(u, probs = c(0.2, 0.4, 0.6, 0.8, 1), na.rm = TRUE))))))
#[1] 1 1 2 2 3 3 4 4 5 5 1 1 2 2 3 3 4 4 5 5
注意:根据OP提出的quantile
问题进行回答