计算自上次活动以来的经过时间 [英] Calculate elapsed time since last event
问题描述
我有一个包含多个主题（ id ）的数据框，重复观察（记录在时间 ）。每个时间都可能与事件相关联（事件）。可以通过以下方式生成示例数据框：
I have a data frame that contains multiple subjects (id), with repeated observations (recorded at times time). Each of the times may or may not be associated with an event (event). An example data frame can be generated with:
set.seed(12345)
id <- c(rep(1, 9), rep(2, 9), rep(3, 9))
time <- c(seq(from = 0, to = 96, by = 12),
seq(from = 0, to = 80, by = 10),
seq(from = 0, to = 112, by = 14))
random <- runif(n = 27)
event <- rep(100, 27)
df <- data.frame(cbind(id, time, event, random))
df$event <- ifelse(df$random < 0.55, 0, df$event)
df <- subset(df, select = -c(random))
df$event <- ifelse(df$time == 0, 100, df$event)
我想计算事件之间的时间（ tae [最后一个事件之后的时间]），理想的输出如下所示：
I would like to calculate the time between events (tae [time after the last event]), such that the ideal output would look like:
head(ideal_df)
id time event tae
1 1 0 100 0
2 1 12 100 0
3 1 24 100 0
4 1 36 100 0
5 1 48 0 12
6 1 60 0 24
在fortr a，我使用以下代码来创建 tae 变量：
In fortran, I use the following code to create the tae variable:
IF(EVENT.GT.0) THEN
TEVENT = TIME
TAE = 0
ENDIF
IF(EVENT.EQ.0) THEN
TAE = TIME - TEVENT
ENDIF
在R中，我尝试了一个 ifelse 和 dplyr 解决方案。但是，既不能产生我想要的输出。
In R, I have attempted both an ifelse and dplyr solution. However, neither produce my desired output.
# Calculate the time since last event (using ifelse)
df$tae <- ifelse(df$event >= 0, df$tevent = df$time & df$tae = 0, df$tae = df$time - df$tevent)
Error: unexpected '=' in "df$tae <- ifelse(df$event >= 0, df$tevent ="
# Calculate the time since last event (using dplyr)
res <- df %>%
arrange(id, time) %>%
group_by(id) %>%
mutate(tae = time - lag(time))
res
id time event tae
1 1 0 100 NA
2 1 12 100 12
3 1 24 100 12
4 1 36 100 12
5 1 48 0 12
6 1 60 0 12
显然，这两个都不会产生我想要的输出。看起来好像在 ifelse 函数中赋予变量不太容易被R容忍。我尝试一个 dplyr 解决方案也无法解释事件 v ariable ...
Clearly, neither of these yield my desired output. It appears as though assigning variables within the ifelse function is not well tolerated by R. My attempt at a dplyr solution also fails to account for the event variable...
最后，需要记录下一个事件 tue 之前的时间的另一个变量。如果有人想到如何最好地去做（也许更棘手）的计算，请随时分享。
Lastly, another variable that recorded the time until the next event tue will be needed. If anyone happens to have a thought regarding how best to go about this (perhaps more tricky) calculation, please feel free to share.
有关如何获得其中一个的任何想法这些工作（或替代解决方案）将不胜感激。谢谢！
Any thoughts regarding how to get one of these working (or an alternative solution) would be greatly appreciated. Thanks!
- 在 ID 之间事件变化之间的时间间隔如下所示：
P.S. -- A reproducible example when the interval between events changes within an ID is presented below:
id <- rep(1, 9)
time <- c(0, 10, 22, 33, 45, 57, 66, 79, 92)
event <- c(100, 0, 0, 100, 0, 100, 0, 0, 100)
df <- data.frame(cbind(id, time, event))
head(df)
id time event
1 1 0 100
2 1 10 0
3 1 22 0
4 1 33 100
5 1 45 0
6 1 57 100
推荐答案
方法与 dplyr ：
library(dplyr)
df %>%
mutate(tmpG = cumsum(c(FALSE, as.logical(diff(event))))) %>%
group_by(id) %>%
mutate(tmp_a = c(0, diff(time)) * !event,
tmp_b = c(diff(time), 0) * !event) %>%
group_by(tmpG) %>%
mutate(tae = cumsum(tmp_a),
tbe = rev(cumsum(rev(tmp_b)))) %>%
ungroup() %>%
select(-c(tmp_a, tmp_b, tmpG))
新列包括事件后的时间（ tae ）和活动前的时间（ tbe ）。
The new columns include time after event (tae) and time before event (tbe).
结果：
id time event tae tbe
1 1 0 100 0 0
2 1 12 100 0 0
3 1 24 100 0 0
4 1 36 100 0 0
5 1 48 0 12 48
6 1 60 0 24 36
7 1 72 0 36 24
8 1 84 0 48 12
9 1 96 100 0 0
10 2 0 100 0 0
11 2 12 0 12 24
12 2 24 0 24 12
13 2 36 100 0 0
14 2 48 0 12 48
15 2 60 0 24 36
16 2 72 0 36 24
17 2 84 0 48 12
18 2 96 0 60 0
19 3 0 100 0 0
20 3 12 100 0 0
21 3 24 0 12 24
22 3 36 0 24 12
23 3 48 100 0 0
24 3 60 100 0 0
25 3 72 100 0 0
26 3 84 0 12 12
27 3 96 100 0 0
第二个例子的结果是：
id time event tae tbe
1 1 0 100 0 0
2 1 10 0 10 23
3 1 22 0 22 11
4 1 33 100 0 0
5 1 45 0 12 12
6 1 57 100 0 0
7 1 66 0 9 26
8 1 79 0 22 13
9 1 92 100 0 0
这篇关于计算自上次活动以来的经过时间的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
