Question

How could I avoid for loop calculating xts weighted sum as I am trying to do below:

library(xts)

thetaSum <- function(theta, w=c(1, 1, 1)) {
    sum(coredata(theta)*rev(w))
}

n <- 10
tmpVec <- rep(1, n)
tmpDates <- seq(as.Date("2000-01-01"), length = n, by = "day")
theta <- xts(tmpVec, order.by=tmpDates)

N <- 3
thetaSummed <- xts(rep(NA, n), order.by=tmpDates)

for (i in N:n) {
    thetaTemp <- theta[(i-N+1):i, ]
    thetaSummed[i] <- thetaSum(thetaTemp, w=rep(1, N))
}

thetaSummed

N is a look back period smaller than n.

What are some fast alternatives for for loop?

Answer 1

You can use rollapply.

rollapplyr(theta, width=3, FUN=thetaSum, fill=NA)