计算data.table中几列的z分数的平均值

Question

In R, I have a data table and a character vector with a subset of the data table's column names. I need to compute the z-scores (ie number of standard deviations from the mean) of each column with a specified name, and put the averages of the z-scores in a new column. I found a solution with explicit for-loops (posted below), but this must be a common enough task that some library function could be made to do the work more elegantly. Is there a better way?

Here's my solution:

#! /usr/bin/env RSCRIPT

library(data.table)

# Sample data table.
dt <- data.table(a=1:3, b=c(5, 6, 3), c=2:4)

# List of column names.
cols <- c('a', 'b')

# Convert columns to z-scores, and add each to a new list of vectors.
zscores <- list()
for (colIx in 1:length(cols)) {
  zscores[[colIx]] <- scale(dt[,get(cols[colIx])], center=TRUE, scale=TRUE)
}

# Average corresponding entries of each vector of z-scores.
avg <- numeric(nrow(dt))
for (rowIx in 1:nrow(dt)) {
  avg[rowIx] <- mean(sapply(1:length(cols),
                            function(colIx) {zscores[[colIx]][rowIx]}))
}

# Add new vector to the table, and print out the new table.
dt[,d:=avg]
print(dt)

This gives what you might expect.

   a b c           d
1: 1 5 2 -0.39089105
2: 2 6 3  0.43643578
3: 3 3 4 -0.04554473

Answer 1

scale can be applied to matrix(-like) object, you can get desired output by

> set(dt, NULL, 'd', rowMeans(scale(dt[, cols, with = F])))
> dt
   a b c           d
1: 1 5 2 -0.39089105
2: 2 6 3  0.43643578
3: 3 3 4 -0.04554473