All,
I'm writing a program where I'm attempting to compute several descriptive statistics with only a single pass through the data. Example statistics are mean, stddev, frequency counts, etc. I am, of course, using DATA Step for this program;).
With most stats, I can use sums or at worst a hash table that stores each unique value along with a count. Is there a calculation formula I can use to incrementally calculate the median (or for that matter, any percentile) without having every unique value available in memory or in a table?
Sorry if this is a VERY naive question - I'm more of a programmer than a statistician:).
Thanks,
Tim Stearn