Calculate mean of a range of rows - r

Question

Given a data frame like this:
A <- c(1,2,3,4,NA,6,7,8,9,10,11,12,13,14,15)
B <- c(NA,NA,NA,20,NA,NA,NA,15,NA,NA,NA,NA,11,NA,9)
DF <- data.frame(A, B)
I would like to calculate the mean for a range of values in column A, based on the value in column B. Specifically, every time there is a non-NA value in column B, I would like to calculate the mean of the range of rows 2 above and 2 below in column A.
For example, the first non-NA value in column B is 20. So I would like to calculate the mean of the two rows above (2, 3), two rows below (NA, 6), and the row adjacent (4). So:
mean(2,3,4,NA,6)
Similarly, the next non-NA value in row B is 15. Which would be
mean(6,7,8,9,10)
So, the end result for the entire data frame would be a new column C
DF$C <- c(NA,NA,NA,3.75,NA,NA,NA,8,NA,NA,NA,NA,13,NA,14)

Answer 1

You could try the following.
nona <- !is.na(DF$B)
DF$C <- replace(
DF$B,
nona,
vapply(which(nona), function(i) {
ii <- (i-2):(i+2)
mean(DF$A[ii[ii > 0]], na.rm = TRUE)
}, 1)
)
Here we are finding the non-NA values in column B and then using that vector to set up the indices for the values we want to find the mean for in column A, being careful to remove any negative subscripts that might occur should the first one or two values of column B not be NA. The above code gives the following result for DF.
A B C
1 1 NA NA
2 2 NA NA
3 3 NA NA
4 4 20 3.75
5 NA NA NA
6 6 NA NA
7 7 NA NA
8 8 15 8.00
9 9 NA NA
10 10 NA NA
11 11 NA NA
12 12 NA NA
13 13 11 13.00
14 14 NA NA
15 15 9 14.00

Answer 2

Here is an approach with the zoo package:
library(zoo)
width <- 5 # the observation ± 2
DF$C <- rollapply(DF$A, width, mean, na.rm = TRUE, partial = TRUE)
# when DF$B is NA, assign NA to corresponding DF$C
DF$C[is.na(DF$B)] <- NA
partial = TRUE allows calculating the mean with a partial window at the leading and trailing parts of the DF$A vector where the whole window can't be accommodated (i.e. the first 2 and last 2 values of DF$A where a window of size 5 is not possible).