# Talk:Percentile

## Weighted Median

I changed the formula for weighted median as I believe it was incorrect. The individual weight of the sample should not be a factor in its percentile. I believe it should be 100/S_N(S_n - 1/2) not 100/S_N(S_n - w_n/2) — Preceding unsigned comment added by Howeman (talk • contribs) 02:08, 23 August 2013 (UTC)

Okay. Apparently my change has been reverted. The current formula is wrong. Let's say I have 1000 samples at 1, 2 samples at 2 and 1 sample at thee

data = [1 2 3] weights = [1000 2 1}

s_n = [1000 1002 1003] S_N = 1003

Using the formula, p_n = [49.85 99.8 99.95]

Now, let's saw we want to find the median (P = 0.5). The integer k for which p_k < P < p_{k+1} is 1. Applying the formula, we get that the median is greater than 1. Given that the 1 values dominate overwhelmingly the dataset, the median should be 1.

Could someone provide a source for the current formula? The R documentation gives a different algorithm http://rss.acs.unt.edu/Rdoc/library/aroma.light/html/weightedMedian.html . — Preceding unsigned comment added by Howeman (talk • contribs) 19:46, 23 August 2013 (UTC)

Reverted your edit because it is incorrect. You have correctly identified a problem with the formula for v, not a problem with the formula for p_n. Rework your example with weights of [1 0.002 0.001] and see that your change to 1/2 doesn't fix the problem. The w_n factor is simply necessary; please leave it in. 18.18.14.50 (talk) 13:47, 25 September 2013 (UTC)

Agreed. The change to 1/2 doesn't work either. I think the article should show the algorithm I linked above. It is both correct and sourced. Is there a source for the current algorithm? Howeman (talk) 20:08, 1 October 2013 (UTC)

## Random chatter

This page is currently slated to be merged with "Quantile", but I really think "Percentile" deserves its own page. The concept is important enough that I wanted to look it up, anyway. --Unregistered Wikipedia-reader, 20 Nov 2004

What is the purpose of the note about Persian translation? --Eve Teschlemacher 19:53, 24 February 2007 (UTC)

Which english reader is interested in the persian script!? I cannot see the relevance and therefore would delete this information or move it to a trivia section. -- 89.48.108.176 13:47, 11 March 2007 (UTC)

Both this page as Percentile rank have the same chinese interwiki, chinese site only has an interwiki to Percentile rank, not to this page. What's the right one, or are they both right? 129.125.155.228 07:43, 3 April 2007 (UTC)

Why are percentiles defined as proportions [0 1], only to be discussed as whole numbers in the range [0 100]?

A common use of the term nth percentile seems to be (based on a recent NYTimes article) the portion with scores in the extremal n%. The article was talking about a gifted program taking those in the top fifth or tenth percentile, i.e. the top 1/20 or 1/10. The definition in the article doesn't seem to admit this usage.

There seems to be a mistake in "Alternative methods": How can k ever be 0, since n is at least 1?

Percentiles and percentile ranks are different terms, and need their own page. "Percenitle" is the score needed to achieve a certain "percentile rank." The two are not interchangealbe. —Preceding unsigned comment added by 67.134.205.79 (talk) 19:31, 12 October 2009 (UTC)

I agree with the above, but the definition of "percentile" and "percentile rank" surely need to be consistent, so the two pages should be harmonised. The "percentile" page states that there is no clear definition: a statement which applies to a sample percentile but not, I think, to a population percentile. The discussion could be improved: a reference to a student report identifying an error in a piece of software is not helpful. User:mjuckes —Preceding unsigned comment added by Mnjuckes (talk • contribs) 11:15, 14 April 2010 (UTC)

## Confusion

Not only should the Percentile and Percentile rank articles be "harmonized", we've got to be careful not to mix up the two concepts in the same article! For example, before I changed it just now the article said:

- "
*A*"**percentile**(or centile) is the value of a variable below which a certain percent of observations fall.

then later, in the "linear interpolation" section:

- "
*we define the percentile corresponding to the n-th value as . In this way, for example, if N = 5 the percentile corresponding to the third value is... 50.*"

No, no, no. That's the percent *rank*, not the percentile. For example: Given the numbers 15, 20, 35, 40, and 50, the percent rank of 20 is 30 (by the above formula), and so the 30th percentile is 20 (i.e., the "percentile" is not 30). - dcljr (talk) 04:51, 12 January 2011 (UTC)

## pronounce

Hello, how is *97.72th percentile* be pronounced? *ninety-seventh point seven two percentile* or *ninety-seven point seven twoth percentile* --Diwas (talk) 11:10, 9 June 2011 (UTC)

Ninety-seven point seventy-second percentile. — Preceding unsigned comment added by 67.0.106.196 (talk) 18:46, 15 April 2012 (UTC)

## Yes, it should be merged

It would limit confusion among other Wikipedia users if both of these articles where merged. They both talk of the exact same concept. --Ctyonahl (talk) 20:43, 20 December 2011 (UTC)

## No, they should not be merged

Percentile and quantile are as different as percentages and fractions. Unless you propose to merge percentage with fractions, these should also remain separate.

Quantile should include a short blurb about percentile, and should link to the percentile article. Percentile should mention quantile in the intro and provide a link (currently where quartiles are mentioned). That makes it easy enough to understand, and falls in line with most other sections of the wiki.

(Percentiles are used in much the same way that percentages are... and quantiles are often used in much the same way that small fractions are. While there is insubstantial mathematical separation between the two, there is definitely a usage difference. There is value to making math articles a little more approachable to the lay person, and keeping them separate will help.)