quantile estimation issue

Here is the link and related parts where I am confused. Especially confused what is the relationship between p and h, are they the same thing?

In effect, the methods compute Qp, the estimate for the k-th q-quantile, where p = k/q, from a sample of size N by computing a real valued index h. When h is an integer, the h-th smallest of the N values, xh, is the quantile estimate. Otherwise a rounding or interpolation scheme is used to compute the quantile estimate from h, x⌊h⌋, and x⌈h⌉. (For notation, see floor and ceiling functions).

$q$ refers to partitioning the set of values into $q$ subsets of (nearly) equal sizes. Check Specialized quantiles in your source.

If you look for a specific quantile $p$, you might express it as $p=k/q$, for example "second decile" where $k=2$

if $h$ is an integer, $h$ can be used as it is as an element index. As your ref says, perhaps not in the neatest way I have seen so far:

When h is an integer, the h-th smallest of the N values, $x_h$, is the quantile estimate.

you might be able to (estimate/calculate) a quantile by looking at the $h$-th indexed element in your the sorted list of (sample/real) population values.

In the most basic example that pops in my head, if you have 10 elements and you would like to calculate the 9th-decile, you could sort your 10 elements and then pick the 9th.

There's a mapping between a quantile (real -valued) and $h$. In the text, they decide to have $h$ be either an integer or a float. You might end-up having a real valued $h$. If you're looking for the 95-something-percentile ($p=0.957658$), well, then as your ref suggests, you would need to:

• apply the floor and ceiling function to $h$
• interpolate (in some way) the elements

I think it would have been neater to use $h$ as an integer valued index valued and have some other quantile (real-value) variable rather than have $h$ be either integer or real-valued and adjust the algo accordingly.

• Thanks IcannotFixThis, vote up, so the statement -- a real valued index h, means h is some existing index in (possible sorted) N data values? My previous confusion is the term real valued, it makes me think real number (en.wikipedia.org/wiki/Real_number). :) – Lin Ma Aug 14 '16 at 22:03
• ok, I see the confusion. I have adjusted the answer to tackled this. – IcannotFixThis Aug 15 '16 at 8:54