在 R 的数据表中查找字符串的第一次迭代

Question

我对 R 还很陌生，所以我想弄清楚如何才能做得更好。 我有一个 data.table，它包含两列（Day 和 Sleepstatus）。 我如何根据列 day 找到睡眠和清醒的第一次迭代，并改变另一列以指示人何时开始睡眠（第一行睡眠）和停止睡眠（第一行清醒）。 睡眠持续时间的 rest，该列应显示 NA

天	睡眠状态
1个	睡眠
1个	睡眠
1个	睡眠
1个	苏醒
2个	睡眠
2个	睡眠
2个	睡眠
2个	苏醒

所需 Output

天	睡眠状态	最终状态
1个	睡眠	开始睡眠
1个	睡眠	北美
1个	睡眠	停止睡眠
1个	苏醒	北美
2个	睡眠	开始睡眠
2个	睡眠	北美
2个	睡眠	停止睡眠
2个	苏醒	北美

Answer 1

这是一个潜在的解决方案：

library(data.table)

dt <- data.table::data.table(
          Day = c(1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L),
  SleepStatus = c("Sleeping","Sleeping","Sleeping",
                  "Awake","Sleeping","Sleeping","Sleeping","Awake")
)

dt[, `Final Status` := {ifelse(
  cumsum(SleepStatus != "Sleeping") != shift(cumsum(SleepStatus != "Sleeping"), fill = 0, type = "lag"),
  "Stop Sleep", "Start Sleep")}]
dt[, `Final Status` := {ifelse(
  `Final Status` == shift(`Final Status`, fill = "NA", type = "lag"),
  NA, `Final Status`)}]
dt
#>    Day SleepStatus Final Status
#> 1:   1    Sleeping  Start Sleep
#> 2:   1    Sleeping         <NA>
#> 3:   1    Sleeping         <NA>
#> 4:   1       Awake   Stop Sleep
#> 5:   2    Sleeping  Start Sleep
#> 6:   2    Sleeping         <NA>
#> 7:   2    Sleeping         <NA>
#> 8:   2       Awake   Stop Sleep

---

如果将代码分解成更小的块，代码会更有意义。 我已经使用下面的 tidyverse 函数完成了此操作，因为我觉得它更容易理解，但如果您愿意，我可以将其更改为 data.table 语法。

library(data.table)

dt <- data.table::data.table(
          Day = c(1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L),
  SleepStatus = c("Sleeping","Sleeping","Sleeping",
                  "Awake","Sleeping","Sleeping","Sleeping","Awake")
)

library(tidyverse)
df <- as.data.frame(dt)

# When the Sleepstatus is not "Sleeping", increment the variable by one
df2 <- df %>%
  mutate(Sleeping = cumsum(SleepStatus != "Sleeping"))
df2
#>   Day SleepStatus Sleeping
#> 1   1    Sleeping        0
#> 2   1    Sleeping        0
#> 3   1    Sleeping        0
#> 4   1       Awake        1
#> 5   2    Sleeping        1
#> 6   2    Sleeping        1
#> 7   2    Sleeping        1
#> 8   2       Awake        2

# If the previous value in "Sleeping" is different to the current value,
# add the "stop sleeping" flag (i.e. show when "Sleeping" changes)
df3 <- df2 %>%
  mutate(Sleep_label = ifelse(Sleeping != lag(Sleeping, default = 0), "Stop sleeping", "Start sleeping"))
df3
#>   Day SleepStatus Sleeping    Sleep_label
#> 1   1    Sleeping        0 Start sleeping
#> 2   1    Sleeping        0 Start sleeping
#> 3   1    Sleeping        0 Start sleeping
#> 4   1       Awake        1  Stop sleeping
#> 5   2    Sleeping        1 Start sleeping
#> 6   2    Sleeping        1 Start sleeping
#> 7   2    Sleeping        1 Start sleeping
#> 8   2       Awake        2  Stop sleeping

# Then, if the value in Sleep_label is equal to the previous label,
# change it to NA
df4 <- df3 %>%
  mutate(Final_status = ifelse(Sleep_label == lag(Sleep_label, default = "NA"), NA, Sleep_label))
df4
#>   Day SleepStatus Sleeping    Sleep_label   Final_status
#> 1   1    Sleeping        0 Start sleeping Start sleeping
#> 2   1    Sleeping        0 Start sleeping           <NA>
#> 3   1    Sleeping        0 Start sleeping           <NA>
#> 4   1       Awake        1  Stop sleeping  Stop sleeping
#> 5   2    Sleeping        1 Start sleeping Start sleeping
#> 6   2    Sleeping        1 Start sleeping           <NA>
#> 7   2    Sleeping        1 Start sleeping           <NA>
#> 8   2       Awake        2  Stop sleeping  Stop sleeping

^{由reprex package (v2.0.1) 创建于 2022-05-20}

那有意义吗？ 还是我只是让事情变得更混乱了？

Answer 2

在 Base R 中，您可以执行以下操作：

x <- dt$SleepStatus
is.na(x) <- -cumsum(c(1,head(rle(x)$lengths,-1)))
dt$final_status <- c(Sleeping = 'Start Sleep', Awake = 'Stop Sleep')[x]
dt

  Day SleepStatus final_status
1   1    Sleeping  Start Sleep
2   1    Sleeping         <NA>
3   1    Sleeping         <NA>
4   1       Awake   Stop Sleep
5   2    Sleeping  Start Sleep
6   2    Sleeping         <NA>
7   2    Sleeping         <NA>
8   2       Awake   Stop Sleep