按ID和序列对R中的表进行分组，没有间隙

Question

我有一张虚构的医院数据表，需要将出院日期替换为（不存在的）人员进行医院转院时的最终出院日期。

rows <- sort(c(which(data$TRANSFER_NUM != 0), which(data$TRANSFER_NUM == 1)-1))
subset <- data[rows,]

令人讨厌的是，有些人可以为不同的事件进行多次转移，即

ID DISCHARGE_DATE FILE_SEQUENCE TRANSFER_NUM 1992-12-04 3360 0 1993-02-11 3361 1 1993-03-10 3362 2 1993-11-25 3363 3 乙 1987-05-15 3419 0 乙 1987-05-19 3420 1 B 1990-02-03 3473 0 B 1990-02-05 3474 1

这意味着

ddply(subset, "ID", mutate, max=max(DISCHARGE_DATE))

会给 B 带来错误的结果，而正确的结果应该是：

ID DISCHARGE_DATE FILE_SEQUENCE TRANSFER_NUM NEW_DISCHARGE_DATE 1992-12-04 3360 0 1993-11-25 1993-02-11 3361 1 1993-11-25 1993-03-10 3362 2 1993-11-25 1993-11-25 3363 3 1993-11-25 B 1987-05-15 3419 0 1987-05-19 B 1987-05-19 3420 1 1987-05-19 B 1990-02-03 3473 0 1990-02-05 B 1990-02-05 3474 1 1990-02-05

我想一些额外的分组可能会有所帮助，也许是这样的：

ID DISCHARGE_DATE FILE_SEQUENCE TRANSFER_NUM GROUP NEW_DISCHARGE_DATE A 1992-12-04 3360 0 1 1993-11-25 1993-02-11 3361 1 1 1993-11-25 1993-03-10 3362 2 1 1993-11-25 A 1993-11-25 3363 3 1 1993-11-25 B 1987-05-15 3419 0 1 1987-05-19 B 1987-05-19 3420 1 1 1987-05-19 B 1990-02-03 3473 0 2 1990-02-05 B 1990-02-05 3474 1 2 1990-02-05

任何帮助将不胜感激！

Answer 1

你是对的，你需要一个中间分组列。这是嵌套的ddply：

ddply(
  ddply(df, "ID", mutate, GROUP=cumsum(c(0, diff(TRANSFER_NUM) < 0))),
  c("ID", "GROUP"),
  mutate, DISCHARGE_NEW=max(as.character(DISCHARGE_DATE))
)
#   ID DISCHARGE_DATE FILE_SEQUENCE TRANSFER_NUM GROUP DISCHARGE_NEW
# 1  A     1992-12-04          3360            0     0    1993-11-25
# 2  A     1993-02-11          3361            1     0    1993-11-25
# 3  A     1993-03-10          3362            2     0    1993-11-25
# 4  A     1993-11-25          3363            3     0    1993-11-25
# 5  B     1987-05-15          3419            0     0    1987-05-19
# 6  B     1987-05-19          3420            1     0    1987-05-19
# 7  B     1990-02-03          3473            0     1    1990-02-05
# 8  B     1990-02-05          3474            1     1    1990-02-05

Answer 2

尝试：

ddply(subset, .(ID,grp=c(0,cumsum(diff(subset$TRANSFER_NUM)-1))), mutate, max=max(DISCHARGE_DATE))

它确实假设 TRANSFER_NUM 是连续的，即 1:x

根据评论，这是我得到的结果：

subset<-read.table(text="ID     DISCHARGE_DATE   FILE_SEQUENCE   TRANSFER_NUM
A      1992-12-04       3360            0
A      1993-02-11       3361            1
A      1993-03-10       3362            2
A      1993-11-25       3363            3
B      1987-05-15       3419            0
B      1987-05-19       3420            1
B      1990-02-03       3473            0
B      1990-02-05       3474            1",header=T)

subset$DISCHARGE_DATE<-as.Date(subset$DISCHARGE_DATE)

ddply(subset, .(ID,grp=c(0,cumsum(diff(subset$TRANSFER_NUM)-1))), mutate, max=max(DISCHARGE_DATE))

  grp ID DISCHARGE_DATE FILE_SEQUENCE TRANSFER_NUM        max
1   0  A     1992-12-04          3360            0 1993-11-25
2   0  A     1993-02-11          3361            1 1993-11-25
3   0  A     1993-03-10          3362            2 1993-11-25
4   0  A     1993-11-25          3363            3 1993-11-25
5  -6  B     1990-02-03          3473            0 1990-02-05
6  -6  B     1990-02-05          3474            1 1990-02-05
7  -4  B     1987-05-15          3419            0 1987-05-19
8  -4  B     1987-05-19          3420            1 1987-05-19

如果grp per ID的子顺序有问题，那么只需更改grp定义前面的符号即可：

ddply(subset, .(ID,grp=-c(0,cumsum(diff(subset$TRANSFER_NUM)-1))), mutate, max=max(DISCHARGE_DATE))

  grp ID DISCHARGE_DATE FILE_SEQUENCE TRANSFER_NUM        max
1   0  A     1992-12-04          3360            0 1993-11-25
2   0  A     1993-02-11          3361            1 1993-11-25
3   0  A     1993-03-10          3362            2 1993-11-25
4   0  A     1993-11-25          3363            3 1993-11-25
5   4  B     1987-05-15          3419            0 1987-05-19
6   4  B     1987-05-19          3420            1 1987-05-19
7   6  B     1990-02-03          3473            0 1990-02-05
8   6  B     1990-02-05          3474            1 1990-02-05