Calculating summaries from arrays without loops

applying-arrays.R

set.seed(1)
n <- 1000

matrices <- replicate(n, matrix(runif(24 ^ 2), nrow = 24))
str(matrices)
#>  num [1:24, 1:24, 1:1000] 0.266 0.372 0.573 0.908 0.202 ...

f <- function(x) {
  sum_x <- sum(x)
  if (sum_x == 0)
    return(0)
  
  p <- x / sum_x
  p[p == 0] <- 1
  
  -sum(p * log2(p))
}

g <- function(x, i) {
  apply(x, i, sum) / sum(x) * 2 ^ apply(x, i, f)
}

result <- apply(matrices, 3, g, 2)
str(result)
#>  num [1:24, 1:1000] 0.926 0.867 0.849 0.89 0.853 ...

#' Created on 2018-03-06 by the [reprex package](http://reprex.tidyverse.org) (v0.2.0).

Ah, yeah I see. Like you've done, just replacing the NAs in result with zeroes will produce the correct result; but still, it's probably cleaner to just deal with that in the f function, too, so that we never divide by zero. If we add a check to see if sum(x) == 0 in f and return 0 if so, we get the same result -- and don't have to do any clean up after!

I'm happy to help! It's been interesting refreshing my array manipulations, since I normally just work with data frames and lists.