user235852 user235852 - 3 months ago 5
R Question

Row sums over columns with a certain pattern in their name

I have a data.table like this

dput(DT)
structure(list(ref = c(3L, 3L, 3L, 3L), nb = 12:15, i1 = c(3.1e-05,
0.044495, 0.82244, 0.322291), i2 = c(0.000183, 0.155732, 0.873416,
0.648545), i3 = c(0.000824, 0.533939, 0.838542, 0.990648), i4 = c(0.044495,
0.82244, 0.322291, 0.393595)), .Names = c("ref", "nb", "i1",
"i2", "i3", "i4"), row.names = c(NA, -4L), class = c("data.table",
"data.frame"), .internal.selfref = <pointer: 0x0000000000320788>)

DT
# ref nb i1 i2 i3 i4
# 1: 3 12 0.000031 0.000183 0.000824 0.044495
# 2: 3 13 0.044495 0.155732 0.533939 0.822440
# 3: 3 14 0.822440 0.873416 0.838542 0.322291
# 4: 3 15 0.322291 0.648545 0.990648 0.393595


Now I want to calculate rows sums, but only including columns which start with an "i" ("i1", "i2", etc)

I have used
grep
to create a vector of the column names to be summed:

listCol <- colnames(DT)[grep("i", colnames(DT))]
listCol
# [1] "i1" "i2" "i3" "i4"


Then I have tried to loop over columns:

DT$sum <- rep.int(0, nrow(DT))
for (i in listCol){
DT$sum = DT$sum + DT[ , get(i)]
}


...which gives the desired output:

DT
# ref nb i1 i2 i3 i4 sum
# 1: 3 12 0.000031 0.000183 0.000824 0.044495 0.045533
# 2: 3 13 0.044495 0.155732 0.533939 0.822440 1.556606
# 3: 3 14 0.822440 0.873416 0.838542 0.322291 2.856689
# 4: 3 15 0.322291 0.648545 0.990648 0.393595 2.355079


How can I improve my code? Many thanks!




This sub question include partially the answer to the previous one :

How to avoid this kind of strange notation :

myrowMeans = function (x){
rowMeans(x, na.rm = TRUE)
}
DT[ , var := myrowMeans(.SD-myrowMeans(.SD)^2), .SDcols = grep("i", colnames(DT))]

Answer

You may also try with Reduce

 DT[, Sum := Reduce(`+`, .SD), .SDcols=listCol][]
 #   ref nb       i1       i2       i3       i4      Sum
 #1:   3 12 0.000031 0.000183 0.000824 0.044495 0.045533
 #2:   3 13 0.044495 0.155732 0.533939 0.822440 1.556606
 #3:   3 14 0.822440 0.873416 0.838542 0.322291 2.856689
 #4:   3 15 0.322291 0.648545 0.990648 0.393595 2.355079

NOTE: If there are "NA" values, it should be replaced with '0' before Reduce i.e.

 DT[, Sum := Reduce(`+`, lapply(.SD, function(x) replace(x, 
                    which(is.na(x)), 0))), .SDcols=listCol][]