在R中缩放data.table列的子集

Question

I would like to scale a subset of columns in my data.table . There are many of these that I would like to scale so i want to avoid specifying them all by name. The columns that are not being scaled, I would just like to return as is. Here is what I was hoping would work but it does not:

require(data.table)
x = data.table(id=1:10, a=sample(1:10,10), b=sample(1:10,10), c=sample(1:10,10))
> dput(x)
structure(list(id = 1:10, a = c(1L, 6L, 10L, 7L, 5L, 3L, 2L, 
4L, 9L, 8L), b = c(4L, 9L, 5L, 7L, 6L, 1L, 8L, 10L, 3L, 2L), 
    c = c(2L, 7L, 5L, 6L, 4L, 1L, 10L, 9L, 8L, 3L)), .Names = c("id", 
"a", "b", "c"), row.names = c(NA, -10L), class = c("data.table", 
"data.frame"), .internal.selfref = <pointer: 0x1a85d088>)

sx = x[,c(id, lapply(.SD, function(v) as.vector(scale(v)))), .SDcols = colnames(x)[2:4]]
   Error in eval(expr, envir, enclos) : object 'id' not found

Any suggestions?

Answer 1

You could also assign by reference in a copy of the data table

sc <- names(x)[2:4]

sx <- copy(x)[ , (sc) := as.data.table(scale(.SD)), .SDcols = sc]

scale returns a matrix and iirc data.table doesn't like matrix columns.

Or,

sx <- copy(x)[ , (sc) := lapply(.SD,scale), .SDcols = sc]

[ The brackets around (sc) tell data.table to take the LHS value from the value of the variable in calling scope rather than the column name sc itself. ]

Answer 2

sx = cbind(x[,-(2:4),with=FALSE],data.table(scale(x[,2:4,with=FALSE])))

我怀疑，您的工作流程会更好地融合您的data.table到长格式。