基于条件 R 的观察子集问题

Question

我希望你能帮助我进行子集设置

数据 1

df2 <- structure(
  list(
    record_id = c(1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2), 
    day_count = c(1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20), 
    event = c(0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0)), 
  row.names = c(NA, -40L),
  class = c("tbl_df", "tbl", "data.frame"))

我的目标我想对事件前后的 5 个观察日 (hazard = 1) 和事件前 12 天的 5 个观察日 (hazard = 2) 进行子集标记和标记创建hazard 变量。请参阅下面的输出：

预期输出 1

df2_output <- structure(
  list(
    record_id = c(1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2), 
    day_count = c(1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20), 
    event = c(0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0),
    hazard = c(NA,NA,NA,2,2,2,2,2,NA,NA,NA,NA,NA,NA,1,1,1,1,1,NA,NA,NA,NA,2,2,2,2,2,NA,NA,NA,NA,NA,NA,1,1,1,1,1,NA)),
  row.names = c(NA, -40L),
  class = c("tbl_df", "tbl", "data.frame"))

尝试 所以我尝试了以下有效的代码

test_df2 <- df2 %>%
  mutate(hazard = case_when(
    (day_count <= df2$day_count[df2$event == 1]) & (day_count > ((df2$day_count[df2$event == 1]) -5)) ~ 1,
    (day_count <= ((df2$day_count[df2$event == 1]) -11)) & (day_count > ((df2$day_count[df2$event == 1]) -16)) ~ 2
    )) %>%
  view()

问题但是，当我在我的主数据集中尝试这个类似的想法时，我收到了以下错误：

Error: Problem with `mutate()` input `hazard`.
x Input `hazard` can't be recycled to size 553.
ℹ Input `hazard` is `case_when(...)`.
ℹ Input `hazard` must be size 553 or 1, not 82.
ℹ The error occurred in group 30: record_id = 120001.
Run `rlang::last_error()` to see where the error occurred.
In addition: There were 50 or more warnings (use warnings() to see the first 50)

我已经确保删除无法充分回顾的事件

我也尝试只选择 1 个记录 ID 来测试代码，但收到以下错误：

4: In day_count > ((case_series_analysis$day_count[case_series_analysis$te_yn ==  :
  longer object length is not a multiple of shorter object length

有人知道吗？

体重

Answer 1

您可以对测试值使用滚动总和。

library(zoo)
library(dplyr)
test_df2 <- df2 %>%
      group_by(record_id) %>%
      mutate( hazard = case_when(
            rollapply(event, 5, sum, partial = TRUE, align = "left") >0 ~ 1,
            rollapply(event, list(11:15), sum, partial = TRUE, fill = 0) >0 ~ 2
      )
      ) %>%
      ungroup()

identical(test_df2, df2_output)
# [1] TRUE

添加了group_by()，因此一条记录中的效果不会影响另一条记录。由于case_when() 会在满足条件后停止，因此在一个事件前 3 天和一秒前 13 天将被编码为 1。

感谢@g-grothendieck 改进了第二个滚动求和的代码。