Cumulative sum in R by group and start over when sum of values in group larger than maximum value

Question

The function below groups values in a vector based on whether the cumulative sum has reached a certain max value and then starts over.

cs_group <- function(x, threshold) {
  cumsum <- 0
  group <- 1
  result <- numeric()
  for (i in 1:length(x)) {
    cumsum <- cumsum + x[i]
    if (cumsum > threshold) {
      group <- group + 1
      cumsum <- x[i]
    }
    result = c(result, group)
  }
  return (result)
}

Example

The max value in the example is 10. The first group only included 9; because summing it with the next value would result in a sum of 12. The next group includes 3, 2, 2 (+8 would result in a value higher then 10).

test <- c(9, 3, 2, 2, 8, 5, 4, 9, 1)
cs_group(test, 10)
[1] 1 2 2 2 3 4 4 5 5

However, I prefer to include in each group the value that results in the cumulative sum to be higher than the maximum value of 10.

Ideal result:

[1] 1 1 2 2 2 3 3 3 4

Answer 1

You can write your own custom function or use the code written by others.

I had the exact same problem few days back and this has been included in the MESS package.

devtools::install_github("ekstroem/MESS")
MESS::cumsumbinning(test, 10, cutwhenpassed = TRUE)
#[1] 1 1 2 2 2 3 3 3 4

Answer 2

One purrr approach could be:

cumsum(c(FALSE, diff(accumulate(test, ~ ifelse(.x >= 10, .y, .x + .y))) <= 0))

[1] 0 0 1 1 1 2 2 2 3

Answer 3

For your purpose, your cs_group can be written like below (if I understand the logic behind in a correct way):

cs_group <- function(x, threshold) {
  group <- 1
  r <- c()
  repeat {
    if (length(x)==0) break
    cnt <- (idx <- max(which(cumsum(x) <= threshold)))+ifelse(idx==length(x),0,1)
    r <- c(r,rep(group, cnt))
    x <- x[-(1:cnt)]
    group <- group + 1
  }
  r
}

such that

test <- c(9, 3, 2, 2, 8, 5, 4, 9, 1)
> cs_group(test, 10)
[1] 1 1 2 2 2 3 3 3 4