在分组 tibble 中应用函数时出错

Question

我正在尝试创建一个函数，该函数获取当前余额，然后每月向后计算，以根据每个月的增加和损失计算上个月末的余额。余额需要在变量组合中计算。我想要的输出看起来像这样 - month_end_balance 是我想要函数输出的。我拥有的其他一切。

type	service	month_starting	add	loss	current_balance	month_end_balance
A	Luxury	12/1/21	2	1	20	20
A	Luxury	11/1/21	4	7	NA	19
A	Luxury	10/1/21	0	0	NA	22
B	Economy	12/1/21	2	8	50	50
B	Economy	11/1/21	4	2	NA	56
B	Economy	10/1/21	0	0	NA	54

我创建了以下函数，它适用于未分组的数据。

running_balance_4 <- function(current_balance, add, loss) {
  out <- rep(NA, length(current_balance))
  out[[1]] <- current_balance[[1]]
  for (i in 2:(length(current_balance))) {
    out[[i]] <-  out[[(i-1)]] - add[[(i-1)]] + loss[[(i-1)]]
  }
  out
}

但我无法在每个组中应用它。这可能只是语法问题。

df %>%
  group_by(type, service) %>%
  arrange(type, service, desc(month_starting)) %>%
  group_modify(running_balance_4(current_balance, add, loss))

感谢任何有关语法和/或函数本身的帮助。

更新：当我尝试运行它时，我收到以下错误消息：找不到对象'current_balance'。所以我认为除了函数的任何问题之外，还可能存在语法错误。

Answer 1

我在 cmets 中看到您的功能正常工作。但是，我认为您可能希望看到使用内置 R 函数而不是应用分组函数的解决方案。

library(tibble)
library(dplyr)

df <- tribble(
  ~type,  ~service, ~month_starting,  ~add, ~loss,  ~current_balance,
  "A", "Luxury",   "12/1/21", 2,  1,  20,
  "A", "Luxury",   "11/1/21", 4,  7,  NA,
  "A", "Luxury",   "10/1/21", 0,  0,  NA,
  "B", "Economy",  "12/1/21", 2,  8,  50,
  "B", "Economy",  "11/1/21", 4,  2,  NA,
  "B", "Economy",  "10/1/21", 0,  0,  NA 
)

df %>% 
  # Two temporary columns to calculate with.
  mutate(
    # Replace current balance with 0 to work with cumulative sum.
    c_balance = coalesce(current_balance, 0),
    # Add the loss and subtract the add since we are working backwards.
    monthly = c_balance + loss - add
  ) %>%
  arrange(type, service, desc(month_starting)) %>%
  group_by(type, service) %>%
  # Taking the lag will put NA on the first element (the rows with current_balance)
  # cumsum is a built in cumulative sum
  mutate(monthly = lag(cumsum(monthly))) %>%
  ungroup() %>%
  mutate(month_end_balance = pmax(current_balance, monthly, na.rm = T))  %>%
  select(-c_balance, -monthly)

它根据current_balance 产生差异，使用cumsum 对该差异进行累积求和，然后使用lag 排列观察结果。然后可以通过current_balance 和滞后变量之间的最大值找到您所追求的输出，因为您不想要的所有值都是NA。

# A tibble: 6 × 7
  type  service month_starting   add  loss current_balance month_end_balance
  <chr> <chr>   <chr>          <dbl> <dbl>           <dbl>             <dbl>
1 A     Luxury  12/1/21            2     1              20                20
2 A     Luxury  11/1/21            4     7              NA                19
3 A     Luxury  10/1/21            0     0              NA                22
4 B     Economy 12/1/21            2     8              50                50
5 B     Economy 11/1/21            4     2              NA                56
6 B     Economy 10/1/21            0     0              NA                54