1. 8
  1.  

  2. 3

    If you’re interested in percentile approximations (mentioned in the very last section of the article), I’d also recommend “biased quantile distributions”, which works by asking you to decide which percentiles are interesting ahead of time (like 0.5, 0.9, 0.99), and then keeps only the samples that ensure it can approximate that answer within an error range. I use that in crow-metrics and have been pleased with the tradeoff.

    1. 1

      Mean, mode, and median are all kinds of average. While this article makes a lot of sense, the fact that it conflates mean and average (to the point of saying things like a median and average will give different answers) makes it quite difficult to read.