generate Zipf distributed data as described in http://arxiv.org/abs/1407.7135

zipf.R

# gnerate Zipf distributed data as described in
#  http://arxiv.org/abs/1407.7135
#  Zipf's law arises naturally in structured, high-dimensional data
#  Laurence Aitchison, Nicola Corradi, Peter E. Latham

n <- 2**22
k <- 20
b <- rnorm(k, 1, 0.2)
data <- replicate(n, {z <- runif(1)
                      p <- (z^b) / (z^b + (1-z)^b)
                      sum((runif(k) < p) * (2**(0:(k-1)))) })
tbl <- sort(table(data), decreasing=TRUE)
plot(log2(1:length(tbl)), log2(tbl))