When looking into statistics for weather data I'm given, usually I get mean, P10, P50 and P90. I would like to name the Pxs in a more friendly way, for the sake of reporting to non-technical people.

In my particular case, the higher percentiles are bad deal - higher weather will imply more costs due to downtime. So after defining clearly what it means statistically, I just give P90 the name "risk" and use that in the text.

Now, how would you call P10? Things like "opportunity" or "profit" come to my mind but I don't really like them in this context.


I advise very strongly against this.

First, if you give the name risk to P90 you are using a very general name, already in broad use in the statistical sciences, for a very specific term.

What then do you call "risk"? You are flying against a mighty wind: it is already a common word with a relevant meaning. People already understand risk broadly even if they are not well informed quantitatively: for example, people may get the relative risks of being shot by small children compared with being shot by terrorists the wrong way round.

Second, this seems too pessimistic. Many people understand 10, 50, 90%, so use those ideas directly. Showing a graph and using a simple numeric example to illustrate are obvious possibilities.

Third, but also important, better informed people will just be puzzled by your not using standard terms. I am an example: I use meteorological and climatological data and am not a statistician, but I understand mean and quantiles.

Whatever you do, some people won't understand, but inventing non-standard terminology (or adopting new meanings for standard terms or common words) isn't the way out.

  • 1
    $\begingroup$ I see your point and I like the idea of not underestimating people's understanding. I'll stick to 10, 50 and 90 percentile. It's short and fits well within the text. In the definitions, a brief explanation will be given anyway. $\endgroup$ – Raf Dec 5 '17 at 9:47
  • $\begingroup$ "People already understand risk broadly even if they are not well informed quantitatively" Not so sure about that one. Quants often use risk to denote a higher standard deviation (=more chances to lose but also more chances to win) whereas in everyday life, we use the term risk to describe situations where there is some chance of losing without necessarily associated chances of winning. I agree though that relabeling P90 as risk is a bad idea, it will add to the existing confusion. $\endgroup$ – David Ernst Dec 5 '17 at 12:49
  • $\begingroup$ @DavidErnst I agree: across disciplines and groups "risk" doesn't mean exactly the same thing. People in finance are welcome to the ugly word "quant" and I suppose they mostly need to be understood by each other.... In this context people, as identified by the OP, means "non-technical people" being exposed to environmental data. Even in that area lower percentiles can signal major problems: a low temperature or a low rainfall or a low windspeed can mean transport problems, crop failures, less energy, etc. so I don't quite agree with the OP's characterisation either. $\endgroup$ – Nick Cox Dec 5 '17 at 13:10
  • 1
    $\begingroup$ @NickCox just to add context, my concern is extreme wave heights and marine transportation. So extreme low values are not a problem. Usually P50 is used for planning and budget. So if weather is good, i.e., no large waves, there's no downtime and the opportunity to increase profit. But if a storm hit, the bill will not be small. So P90 is always bad and imply in loss of profit, while P10 is always good. But anyway, I agree with you and as mentioned will stick to Pn in the reports. If that is not clear to someone, they're welcome to ask =) $\endgroup$ – Raf Dec 7 '17 at 10:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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