CODEWITHSUNDEEP
r
Anna
Anna

Reputation: 895

How to perform a cumulative sum with unique IDs only?

I have the following data frame:

d<-data.frame(Day=c(1, 1, 1, 1, 1, 1, 2), ID=c("A", "B", "C", "D", "A", "B", "B"), Value=c(1, 2, 3, 4, 5, 6, 7))

On each day, I would like a cumulative sum of unique values, taking only the most recent value for an entry that repeats. My expected output is as follows:

d<-data.frame(Day=c(1, 1, 1, 1, 1, 1, 2), ID=c("A", "B", "C", "D", "A", "B", "B"), Value=c(1, 2, 3, 4, 5, 6, 7), Sum=c(1, 3, 6, 10, 14, 18, 7))

  Day ID Value Sum
1   1  A     1   1
2   1  B     2   3
3   1  C     3   6
4   1  D     4  10
5   1  A     5  14
6   1  B     6  18
7   2  B     7   7

where the 5th entry adds up values 2, 3, 4, 5 (because A repeats) and the 6th entry adds up values 3, 4, 5, and 6 (because both A and B repeat). The 7th entry restarts because it is a new day.

I don't think I can use cumsum() as it only accepts 1 parameter. I also don't want to keep a counter for each ID as I may have up to 100 unique IDs per day.

Any hints or help would be appreciated! Thank you!

Upvotes: 11

Views: 392

Answers (1)

Frank
Frank

Reputation: 66819

You can difference the values by ID and Day and then use cumsum:

library(data.table)
setDT(d)
d[, v_eff := Value - shift(Value, fill=0), by=.(Day, ID)]
d[, s := cumsum(v_eff), by=Day]

   Day ID Value Sum v_eff  s
1:   1  A     1   1     1  1
2:   1  B     2   3     2  3
3:   1  C     3   6     3  6
4:   1  D     4  10     4 10
5:   1  A     5  14     4 14
6:   1  B     6  18     4 18
7:   2  B     7   7     7  7

Base R analogue...

d$v_eff <- with(d, ave(Value, Day, ID, FUN = function(x) c(x[1], diff(x)) ))
d$s <- with(d, ave(v_eff, Day, FUN = cumsum))

  Day ID Value Sum v_eff  s
1   1  A     1   1     1  1
2   1  B     2   3     2  3
3   1  C     3   6     3  6
4   1  D     4  10     4 10
5   1  A     5  14     4 14
6   1  B     6  18     4 18
7   2  B     7   7     7  7

Upvotes: 5

Related Questions