Joe wrote:

Has anyone written a percentile rank algorithm that they can share or

know of a good resource that defines the algorithm. I'm starting a

project that involves percentile ranks, logistic and linear regression.

If not c# any language or pseudo code would be very useful. Thank you.

Joe

--

Sent via .NET Newsgroups

http://www.dotnetnewsgroups.com

Forgive me if I'm wrong but isn't percentile rank just counting how many

of the values are below or equal the specific value you're looking at in

terms of percentage of the entire list of values ?

Ie. if you have a value of 50, and there are 75 values in the list, and

30 of those are 50 or below, wouldn't this mean that the percentile rank

of 50 in this list is (30/75)*100 = 40%?

I'd assume sorting the list, then running through it and assigning

percentile ranks would be the best way. You'd have to handle duplicate

values though:

1, 1, 2, 2, 3, 3, 4

would have percentile ranks of:

25%, 25%, 50%, 50%, 75%, 75%, 100%

(if I'm not wrong about what a percentile rank is)

pseudo-code for the above:

- sort list

- loop through all elements from start to finish (lowest to highest)

- find all duplicates of the value you're looking at

- calculate percentile rank as number of values below the value

(index of the loop), plus the number of duplicates, divided by

the number of elements in the list

- skip past duplicates

--

Lasse Vågsæther Karlsen

http://usinglvkblog.blogspot.com/
mailto:la***@vkarlsen.no

PGP KeyID: 0x2A42A1C2