R: Is there a column version for "rowsum", but to compute the mean and not just the sum?

Question

I have a matrix, with columns grouped according to a grouping variable:

x <- matrix(sample(1:10,20,T),ncol=4)
     [,1] [,2] [,3] [,4]
[1,]    7    8    5    3
[2,]    7    5    4    7
[3,]    7    1    9    3
[4,]    4    8    8    8
[5,]    9    9    1    5

group <- sample(1:2, 4, TRUE)
[1] 1 2 1 2

What is the most elegant way to compute the mean of each row grouped by the column grouping variable? The result for the example shown should be the 5 by 2 matrix:

       1   2
[1,] 6.0 5.5
[2,] 5.5 6.0
[3,] 8.0 2.0
[4,] 6.0 8.0
[5,] 5.0 7.0

I looked at the rowsum command, but that sums the rows by the grouping variable, and does not have a readymade option to compute the mean.

I would greatly appreciate any help.

Answer 1

We can use split.default on a data.frame by the 'group' to split by columns and then do the rowMeans on the list of the data.frames

sapply(split.default(as.data.frame(x), group), rowMeans)
#      1   2
#[1,] 6.0 5.5
#[2,] 5.5 6.0
#[3,] 8.0 2.0
#[4,] 6.0 8.0
#[5,] 5.0 7.0

data

x <- structure(c(7, 7, 7, 4, 9, 8, 5, 1, 8, 9, 5, 4, 9, 8, 1, 3, 7, 
       3, 8, 5), .Dim = 5:4)
group <- c(1, 2, 1, 2)