Weighted average across multiple matrices - item by item
I have "mylist" - a list of matrices of the same size:
mylist <- vector("list", 5)
set.seed(123)
for(i in 1:5){
mylist[[i]] <- matrix(rnorm(9), nrow = 3)
}
I also have a vector of weights "mywgts" - the same length as "mylist"
mywgts <- c(0.8, 0.9, 1, 1.1, 1.2)
I need to calculate the weighted average of these matrices - element by element. The result will be a 3-by-3 matrix, where the first element is:
mylist[[1]][1,1]*mywgts[1] + mylist[[2]][1,1]*mywgts[2] +
mylist[[3]][1,1]*mywgts[3] + mylist[[4]][1,1]*mywgts[4] +
mylist[[5]][1,1]*mywgts[5]
I know how to do this by iterating over all the elements of the matrix. But I am looking for a more economical / elegant R-shaped solution. Also - the actual length of "mylist" is not known in advance.
Thanks for any hint!
source to share
You may try
res <- Reduce(`+`,Map(`*`, mylist, mywgts))
res
# [,1] [,2] [,3]
#[1,] 0.6852912 0.2116715 0.7993867
#[2,] -0.8815045 -1.9811868 1.2558095
#[3,] 1.5150166 0.8780412 0.7254080
Map
is a wrapper for mapply
which is a multidimensional version sapply
. The ( *
) function is applied to the matching elements of the first ("mylist") and the second elements ("mywgts"), and then uses them Reduce
to sum the matching elements list
.
If you need mean
, split it by the length of "mylist".
res/length(mylist)
Using OP's computation
mylist[[1]][1,1]*mywgts[1] + mylist[[2]][1,1]*mywgts[2] +
mylist[[3]][1,1]*mywgts[3] + mylist[[4]][1,1]*mywgts[4] +
mylist[[5]][1,1]*mywgts[5]
#[1] 0.6852912
mylist[[1]][1,2]*mywgts[1] + mylist[[2]][1,2]*mywgts[2] +
mylist[[3]][1,2]*mywgts[3] + mylist[[4]][1,2]*mywgts[4] +
mylist[[5]][1,2]*mywgts[5]
#[1] 0.2116715
source to share