This is part 3 of my series on calculating percentiles on streaming data.
In an effort to better understand the Greenwald-Khanna [GK01] algorithm, I created a series of visualizations of the cumulative distribution functions of a randomly-generated, normally-distributed data set with = 0 and = 1, as the number of random numbers increases from 1 to 1,000.
The way to read these visualizations is to find the percentile you are looking for on the y-axis, then trace horizontally to find the vertical line on the chart which intersects this location, then read the value from the x-axis.
From these visualizations, it is quite intuitive and clear how the “resolution” of Greenwald-Khanna increases as decreases, and how the compress operation keeps the number of elements in the summary data set relatively small as increases.
- [GK01] M. Greenwald and S. Khanna. Space-efficient online computation of quantile summaries. In Proceedings of ACM SIGMOD, pages 58–66, 2001.