How to do this in R - r

Question

I have a dataset that looks like this:
groups <- c(1:20)
A <- c(1,3,2,4,2,5,1,6,2,7,3,5,2,6,3,5,1,5,3,4)
B <- c(3,2,4,1,5,2,4,1,3,2,6,1,4,2,5,3,7,1,4,2)
position <- c(2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1)
sample.data <- data.frame(groups,A,B,position)
head(sample.data)
groups A B position
1 1 1 3 2
2 2 3 2 1
3 3 2 4 2
4 4 4 1 1
5 5 2 5 2
6 6 5 2 1
The "position" column always alternates between 2 and 1. I want to do this calculation in R: starting from the first row, if it's in position 1, ignore it. If it starts at 2 (as in this example), then calculate as follows:
Take the first 2 values of column A that are at position 2, average them, then subtract the first value that is at position 1 (in this example: (1+2)/2 - 3 = -1.5). Then repeat the calculation for the next set of values, using the last position 2 value as the starting point, i.e. the next calculation would be (2+2)/2 - 4 = -2.
So basically, in this example, the calculations are done for the values of these sets of groups: 1-2-3, 3-4-5, 5-6-7, etc. (the last value of the previous is the first value of the next set of calculation)
Repeat the calculation until the end. Also do the same for column B.
Since I need the original data frame intact, put the newly calculated values in a new data frame(s), with columns dA and dB corresponding to the calculated values of column A and B, respectively (if not possible then they can be created as separated data frames, and I will extract them into one afterwards).
Desired output (from the example):
dA dB
1 -1.5 1.5
2 -2 3.5
3 -3.5 2.5
4 -4.5 2.5
5 -4.5 2.5
6 -2.5 4

Answer 1

groups <- c(1:20)
A <- c(1,3,2,4,2,5,1,6,2,7,3,5,2,6,3,5,1,5,3,4)
B <- c(3,2,4,1,5,2,4,1,3,2,6,1,4,2,5,3,7,1,4,2)
position <- c(2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1)
sample.data <- data.frame(groups,A,B,position)
start <- match(2, sample.data$position)
twos <- seq(from = start, to = nrow(sample.data), by = 2)
df <-
sapply(c("A", "B"), function(l) {
sapply(twos, function(i) {
mean(sample.data[c(i, i+2), l]) - sample.data[i+1, l]
})
})
df <- setNames(as.data.frame(df), c('dA', 'dB'))

Answer 2

As your values in position always alternate between 1 and 2, you can define an index of odd rows i1 and an index of even rows i2, and do your calculations:
## In case first row has position==1, we add an increment of 1 to the indexes
inc=0
if(sample.data$position[1]==1)
{inc=1}
i1=seq(1+inc,nrow(sample.data),by=2)
i2=seq(2+inc,nrow(sample.data),by=2)
res=data.frame(dA=(lead(sample.data$A[i1])+sample.data$A[i1])/2-sample.data$A[i2],
dB=(lead(sample.data$B[i1])+sample.data$B[i1])/2-sample.data$B[i2]);
This returns:
dA dB
1 -1.5 1.5
2 -2.0 3.5
3 -3.5 2.5
4 -4.5 2.5
5 -4.5 2.5
6 -2.5 4.0
7 -3.5 2.5
8 -3.0 3.0
9 -3.0 4.5
10 NA NA
The last row returns NA, you can remove it if you need.
res=na.omit(res)