I would like to write code to compute within each group, sum of lagged differences as shown in the table below:
ID x rank U R Required Output Value
1 1 1 U1 R1 -
1 1 2 U2 R2 R2-U1
1 1 3 U3 R3 (R3-U2) + (R3-U1)
1 1 4 U4 R4 (R4-U3) + (R4-U2) + (R4-U1)
1 0 5 U5 R5 R5
1 0 6 U6 R6 R6
1 0 7 U7 R7 R7
2 1 1 U8 R8 -
2 1 2 U9 R9 R9-U8
2 1 3 U10 R10 (R10-U9) + (R10 - U8)
2 1 4 U11 R11 (R11-U10) + (R11 - U9) + (R11 - U8)
3 1 1 U12 R12 -
3 0 2 U13 R13 R13
3 0 3 U14 R14 R14
ID is the unique group identifier. x is a bool and depending on its value the required output is either sum of difference with previous values or same period value. "rank" is a rank ordering column and the maximum rank can vary within each group. "U" and "R" are the main columns of interest.
To give a numerical example, I need the following:
ID x rank U R Required Output Value
1 1 1 10 7 -
1 1 2 9 11 1
1 1 3 10 10 1 + 0 = 1
1 1 4 11 13 3+4+3 = 10
1 0 5 7 8 8
1 0 6 8 8 8
1 0 7 5 7 7
2 1 1 3 2 -
2 1 2 9 15 12
2 1 3 13 14 16
2 1 4 1 14 17
3 1 1 12 1 -
3 0 2 14 9 9
3 0 3 1 11 11
R code to generate this table:
ID = c(rep(1,7),rep(2,4),rep(3,3))
x = c(rep(1,4),rep(0,3),rep(1,5),rep(0,2))
rank = c(1:7,1:4,1:3)
U = c(10,9,10,11,7,8,5,3,9,13,1,12,14,1)
R = c(7,11,10,13,8,8,7,2,15,14,14,1,9,11)
dat = cbind(ID,x,rank,U,R)
colnames(dat)=c("ID","x","rank","U","R")
Here's a tidyverse solution:
library(dplyr)
library(tidyr)
dat %>%
as_tibble() %>%
group_by(ID) %>%
mutate(output = ifelse(x, lag(rank) * R - lag(cumsum(U)), R))
Result:
# A tibble: 14 x 6
# Groups: ID [3]
ID x rank U R output
<dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 1 1 1 10 7 NA
2 1 1 2 9 11 1
3 1 1 3 10 10 1
4 1 1 4 11 13 10
5 1 0 5 7 8 8
6 1 0 6 8 8 8
7 1 0 7 5 7 7
8 2 1 1 3 2 NA
9 2 1 2 9 15 12
10 2 1 3 13 14 16
11 2 1 4 1 14 17
12 3 1 1 12 1 NA
13 3 0 2 14 9 9
14 3 0 3 1 11 11