Summing across rows of a data.table for specific columns

Question

I have a large data table (from the package data.table) with over 60 columns (the first three corresponding to factors and the remaining to response variables, in this case

Answer 1

An alternative (data.table) approach would be to store your data in long form. Version 1.8.11 of data.table has fast melt and dcast methods

library(reshape2)
mt <- melt(test, id=1:3,variable.name='Species')

abundance <- mt[,list(abundance = mean(value)),by=list(Zone,quadrat,Species)][, 
                sumAbundance := sum(abundance), by = list(Zone,quadrat)]

Working in long format will take a slight change in thinking, but it may end up being more efficient memory wise (as less internal copying will be involved, and you are referencing a single not multiple elements within every "by" group.)

Answer 2

[ Edited 2020-02-15 to reflect current state of data.table ] In recent versions of data.table rowSums(Abundance[ , 4:6]) works as OP originally expected. Here are some alternatives:

Abundance[, SumAbundance := rowSums(.SD), .SDcols = 4:6]

Also, I didn't check, but I have a suspicion this will be faster, since it will not convert to matrix as rowSums does:

Abundance[, SumAbundance := Reduce(`+`, .SD), .SDcol = 4:6]