I have a data frame like this:

df <- data.frame(a = runif(100, -10, 10),
                 b = runif(100, -10, 10),
                 c = runif(100, -10, 10),
                 d = runif(100, -10, 10),
                 e = runif(100, -10, 10))

Now, I’d like to have a function that extracts n rows with the maximum dispersal.

To explain a little better, imagine a one-dimensional data set like c(1:9). The function I am looking for would be something like fun(3, c(1:9)) (with n=3) and return c(1, 5, 9). fun(5, c(1:9)) would return c(1, 3, 5, 7, 9).

A valid result for fun(3, c(1:10)) would be c(1, 5, 10) or c(1, 6, 9). In that case, the function should randomly select one of the valid outputs.

A while ago, I wrote this:

voronoiFilter <- function(occ, select){
  n <- nrow(occ) - select
  subset <- occ
  dropped <- rep(NA, n)
  for (i in 1:n) {
    v <- voronoi.mosaic(x = subset[,1], y = subset[,2], duplicate = "remove")
    info <- cells(v)
    areas <- unlist(lapply(info,function(x) x$area))
    smallest <- which(areas == min(areas, na.rm = TRUE))
    dropped[i] <- which(occ[,1] == subset[smallest,1] & occ[,2] == subset[smallest,2])
    subset <- subset[-smallest,]
  }
  outVec <- 1:nrow(occ)
  return(outVec[-dropped])
}

where occ is the input data frame and selected is n.

This sometimes works OKish, but often I receive this error:

Warning in hist.default(i, plot = FALSE, freq = TRUE, breaks = seq(0.5,  :
  argument ‘freq’ is not made use of

I think because there are duplicated elements in the data.

Does anybody have another idea, how such a function could look like?