CODEWITHSUNDEEP
r
Spiffy
Spiffy

Reputation: 3

Summing across 2 columns for multiple columns and skipping redundant combinations

I want to find the sums (by row) for values within all unique two column combinations in a matrix. For example I want to know V1+V2, therefore I don't need to know V2+V1 and I also don't need V1+V1, V2+V2 and so on.

Using some of the info I have gathered on here, I've been able to create a loop function that partially solves my problem. It sums the values in columns 1 through 4 paired with the values in columns 2 through 5.

set.seed(3)
mydata<-as.data.frame(matrix(rexp(20, rate=.1), ncol=5))
for (i in 1:4) {
for (j in 2:5) {
Newcolname <- paste0(names(mydata)[i],"_",names(mydata)[j]) 
mydata[[Newcolname]] <- rowSums(mydata[,c(i,j)])
}
}

This code creates what I want but with too much information. I get 16 new columns with sums from columns 1 through 4 paired with columns 2 through 5. This includes redundant information that I'd like to exclude.

Ideally, I'd like to run this function only in cases where j>i so it should create 10 new columns with these sums: V1_V2, V1_V3, V1_V4, V1_V5, V2_V3, V2_4, V2_V5, V3_V4, V3_V5, V4_V5. Thanks for any help. I'm new to creating loop functions.

Upvotes: 0

Views: 38

Answers (1)

Shree
Shree

Reputation: 11150

Here's a way using combn from base R -

combn(1:ncol(mydata), 2, FUN = function(x) {rowSums(mydata[, x])})

Here's the code with naming. I wonder if there's a better way -

# generate combinations only once; to be used for calculation and naming
combos <- combn(1:ncol(mydata), 2)    
# calculate sums for each combo
res <- apply(combos, 2, function(x) {
  rowSums(mydata[, x])
})    
# set names
colnames(res) <- apply(combos, 2, function(x) paste0("V", x, collapse = "_"))    
res

         V1_V2     V1_V3    V1_V4    V1_V5     V2_V3     V2_V4    V2_V5
[1,] 19.348279 17.986973 18.88804 21.19083  2.722721  3.623788  5.92658
[2,]  8.238209  7.295317 48.38643 18.24868  3.232870 44.323986 14.18623
[3,] 35.165599 13.101673 18.15211 24.19188 23.608940 28.659382 34.69914
[4,] 10.141323 14.114612 12.03548 27.52704  4.174245  2.095110 17.58667
         V3_V4     V3_V5     V4_V5
[1,]  2.262482  4.565274  5.466342
[2,] 43.381094 13.243336 54.334452
[3,]  6.595455 12.635218 17.685659
[4,]  6.068398 21.559962 19.480826

Upvotes: 1

Related Questions