R 计数组中的编号，但对于非连续日期重新启动

我有如下数据：

sample <- data.frame(
  group = c("A","A","A","B","B","B"),
  date = c(as.Date("2014-12-31"),
           as.Date("2015-01-31"),
           as.Date("2015-02-28"),
           as.Date("2015-01-31"),
           as.Date("2015-03-31"),
           as.Date("2015-04-30")),
  obs = c(100, 200, 300, 50, 100, 150)
)

我想创建一个第四列，计算组中观察的数量。但是，如果一个月没有紧接着前一个月，我希望重新开始计数。这就是我希望它看起来的样子：

  group       date obs num
1     A 2014-12-31 100   1
2     A 2015-01-31 200   2
3     A 2015-02-28 300   3
4     B 2015-01-31  50   1
5     B 2015-03-31 100   1
6     B 2015-04-30 150   2

到目前为止，我只能得到以下信息：

library(tidyverse)
sample <- sample %>%
  arrange(date) %>%
  group_by(group) %>%
  mutate(num = row_number())

  group       date obs num
1     A 2014-12-31 100   1
2     A 2015-01-31 200   2
3     A 2015-02-28 300   3
4     B 2015-01-31  50   1
5     B 2015-03-31 100   2
6     B 2015-04-30 150   3

---

任何帮助都将不胜感激。我也希望能够做同样的事情，但是使用季度数据而不是月度数据。

我们可以根据“日期”月份的差异创建一个组，如果不等于1，即一个月的差异

library(dplyr)
library(lubridate)
sample %>%
     arrange(group, date) %>% 
     group_by(group, mth = cumsum(c(TRUE, diff(month(date)) != 1))) %>% 
     mutate(num = row_number()) %>%
     ungroup %>%
     select(-mth)
# A tibble: 6 x 4
#  group date         obs   num
#  <fct> <date>     <dbl> <int>
#1 A     2015-01-31   100     1
#2 A     2015-02-28   200     2
#3 A     2015-03-31   300     3
#4 B     2015-01-31    50     1
#5 B     2015-03-31   100     1
#6 B     2015-04-30   150     2

如果年度也需要考虑

library(zoo)
sample %>%
    arrange(group, date) %>% 
    mutate(yearmon = as.yearmon(date)) %>% 
    group_by(group) %>%
    group_by(grp = cumsum(c(TRUE, as.integer(diff(yearmon) * 12)> 1)),
             add = TRUE ) %>% 
    mutate(num = row_number()) %>% 
    ungroup %>% 
    select(-grp, -yearmon)
# A tibble: 6 x 4
#  group date         obs   num
#  <fct> <date>     <dbl> <int>
#1 A     2015-01-31   100     1
#2 A     2015-02-28   200     2
#3 A     2015-03-31   300     3
#4 B     2015-01-31    50     1
#5 B     2015-03-31   100     1
#6 B     2015-04-30   150     2

我们可以使用lubridate:：days_in_month来获取一个月的天数，并将其与当前日期和过去日期的差值进行比较，以创建一个新组。然后我们可以在每个组中分配行号

library(dplyr)

sample %>%
  group_by(group) %>%
  mutate(diff_days = cumsum(as.numeric(date - lag(date, default = first(date))) != 
                     lubridate::days_in_month(date))) %>%
  group_by(diff_days, add = TRUE) %>%
  mutate(num = row_number()) %>%
  ungroup() %>%
  select(-diff_days)


# group  date         obs   num
#  <fct> <date>     <dbl> <int>
#1 A     2014-12-31   100     1
#2 A     2015-01-31   200     2
#3 A     2015-02-28   300     3
#4 B     2015-01-31    50     1
#5 B     2015-03-31   100     1
#6 B     2015-04-30   150     2

---

非常感谢。我意识到我的样本数据过于简化了。并非所有日期都在同一年内。我改变了最初的评论以反映这一点。@H.Z当你说月差时，它是与每天的绝对差异，还是只考虑了年和月

library(dplyr)

sample %>%
  group_by(group) %>%
  mutate(diff_days = cumsum(as.numeric(date - lag(date, default = first(date))) != 
                     lubridate::days_in_month(date))) %>%
  group_by(diff_days, add = TRUE) %>%
  mutate(num = row_number()) %>%
  ungroup() %>%
  select(-diff_days)


# group  date         obs   num
#  <fct> <date>     <dbl> <int>
#1 A     2014-12-31   100     1
#2 A     2015-01-31   200     2
#3 A     2015-02-28   300     3
#4 B     2015-01-31    50     1
#5 B     2015-03-31   100     1
#6 B     2015-04-30   150     2