r - 满足条件时跟踪最后一次观察的组特定变量

Question

我正在尝试构建一个特定于组的变量:

1. 在满足条件之前进行观察。
2. 满足条件时的第一次观察不适用。
3. 对于后续观察，为满足条件时上次观察的值。

下面是 MWE。我正在寻找构建“last_value”的代码。我更喜欢使用 data.table(或 dplyr)来实现可扩展性/性能，但我愿意接受其他建议。谢谢!

library(data.table)

data.table(
    group = c(rep("alpha", 4), rep("beta", 3)),
    time = c(1:4, 1:3),
    condition = c("Yes", "No", "Yes", "No", "No", "Yes", "No"),
    value = 1:7,
    last_value = c(NA, 1, 1, 3, NA, NA, 6))

#    group time condition value last_value
# 1: alpha    1       Yes     1         NA
# 2: alpha    2        No     2          1
# 3: alpha    3       Yes     3          1
# 4: alpha    4        No     4          3
# 5:  beta    1        No     5         NA
# 6:  beta    2       Yes     6         NA
# 7:  beta    3        No     7          6

# last_value is:

#     NA in the 1st row as that is the 1st observation for group "alpha"
#     1 in the 2nd row as the 1st observation is condition = "Yes"
#     1 in the 3rd row as the 1st observation is condition = "Yes"; 2nd observation is condition = "No"
#     3 in the 4rd row as the 3rd observation is condition = "Yes"

#     NA in the 5th row as that is the 1st observation for group "beta"
#     NA in the 6th row as there is no prior observation with condition = "Yes"
#     6 in the 7th row as the 6th observation is condition = "Yes"

Answer 1

这是 dplyr 中使用 cummax 的另一个。

library(dplyr)
df %>%
  group_by(group) %>%
  mutate(last = cummax(row_number() * (condition == "Yes")),
         last = lag(value[replace(last, last == 0, NA)]))


#  group  time condition value last_value  last
#  <fct> <int> <fct>     <int>      <dbl> <int>
#1 alpha     1 Yes           1         NA    NA
#2 alpha     2 No            2          1     1
#3 alpha     3 Yes           3          1     1
#4 alpha     4 No            4          3     3
#5 beta      1 No            5         NA    NA
#6 beta      2 Yes           6         NA    NA
#7 beta      3 No            7          6     6