Say I have data from a sample of 100 experiments. For a variable, I'm interested in showing the value x with 0.9 probability of non-exceedance, which is equivalent to P90 of the sorted sample.

I'm interested in proper names for these parameters:

x is a percentile, the 90-percentile to be specific, or more generally speaking, a quantile.

90 is the rank of x in my sample. Correct?

What would be the proper name for 0.9, or 90%? Also rank? Normalized rank? Strictly speaking it is the probability of non-exceedance, but writing that in a table header is not convenient :) A short but correct name for it would be very welcome.

  • 1
    $\begingroup$ In ordinary usage 'parameter' refers to a population, not a sample. The word 'statistic' is more commonly used for samples. $\endgroup$
    – BruceET
    Jul 25, 2018 at 16:46

1 Answer 1


I would call it the level that is used to calculate the quantiles and percentiles. (I have seen this referred to as a confidence level, but I don't like the insinuation that it would mostly be used for .)

(And yes, since you have 100 observations, the rank of the 90% percentile observation is 90. If you had 200 observations, then the 90% percentile observation would have rank 180.)

  • $\begingroup$ I have also seen it as confidence level and I don't like for the same reason you've mentioned. Level somehow has a too strong resemblance to confidence level so not my preferred option. Calling it rank is definitely wrong then. About normalized rank? if that makes sense... $\endgroup$
    – Raf
    Jul 25, 2018 at 8:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.