在r中的标记列重置的cumsum？

Question

这是我第一次提出问题，请多多包涵。

我的数据集（df）是这样的：

animal   azimuth   south   distance
 pb1      187.561   1       1.992 
 pb1      147.219   1       8.567
 pb1      71.032    0       5.754
 pb1      119.502   1       10.451
 pb2      101.702   1       9.227
 pb2      85.715    0       8.821

我想创建一个额外的列 (df$cumdist) 来增加累积距离，但在每个单独的动物内并且仅当 df$south==1 时。我希望用df$south==0 重置累积和。

这就是我想要的结果（手动完成）：

animal   azimuth   south   distance  cumdist
 pb1      187.561   1       1.992     1.992
 pb1      147.219   1       8.567     10.559 
 pb1      71.032    0       5.754     0 
 pb1      119.502   1       10.451    10.451
 pb2      101.702   1       9.227     9.227 
 pb2      85.715    0       8.821     0

这是我尝试实现 cumsum 的代码：

swim.az$cumdist <- cumsum(ifelse(swim.az$south==1, swim.az$distance, 0))

虽然在df$south==0 时成功停止添加，但不会重置。此外，我知道我需要将其嵌入到 for 循环中，以按动物进行子集。

非常感谢！

Answer 1

我们将 'south' 与 'distance' ('cumdist') 相乘以将 'south' 中的 0 对应的 'distance' 中的值更改为 0，按 'animal' 分组，并通过获取累积值创建的组逻辑向量的总和（south == 0），得到'cumdist'的cumsum，ungroup并删除不需要的列（grp）

library(dplyr)
dfN %>% 
  mutate(cumdist = south * distance) %>%
  group_by(animal, grp = cumsum(south == 0)) %>%
  mutate(cumdist = cumsum(cumdist)) %>%
  ungroup %>%
  select(-grp)
# A tibble: 6 x 5
#  animal azimuth south distance cumdist
#  <chr>    <dbl> <int>    <dbl>   <dbl>
#1 pb1      188.      1     1.99    1.99
#2 pb1      147.      1     8.57   10.6 
#3 pb1       71.0     0     5.75    0   
#4 pb1      120.      1    10.5    10.5 
#5 pb2      102.      1     9.23    9.23
#6 pb2       85.7     0     8.82    0   

---

或base R 的类似方法

with(dfN, ave(distance * south, animal, cumsum(!south), FUN = cumsum))
#[1]  1.992 10.559  0.000 10.451  9.227  0.000

数据

dfN <- structure(list(animal = c("pb1", "pb1", "pb1", "pb1", "pb2", 
"pb2"), azimuth = c(187.561, 147.219, 71.032, 119.502, 101.702, 
85.715), south = c(1L, 1L, 0L, 1L, 1L, 0L), distance = c(1.992, 
8.567, 5.754, 10.451, 9.227, 8.821)), class = "data.frame", 
row.names = c(NA, -6L))

Answer 2

library(data.table)
setDT(df)

df[, cumdist := south*cumsum(distance), .(animal, rleid(south))]

#    animal azimuth south distance cumdist
# 1:    pb1 187.561     1    1.992   1.992
# 2:    pb1 147.219     1    8.567  10.559
# 3:    pb1  71.032     0    5.754   0.000
# 4:    pb1 119.502     1   10.451  10.451
# 5:    pb2 101.702     1    9.227   9.227
# 6:    pb2  85.715     0    8.821   0.000