Thank you Kip.
The 5&simplify function is equally sensitive to all observations while the
quartiles are highly sensititive to a few observations.
The 5&simplify function is analytic while the quartiles are not analytic
functions.
Den 12:19 torsdag den 25. juni 2015 skrev Kip Murray
<[email protected]>:
You may want to try Tukey's five number, seven member, etc summaries,
appropriate for data from distributions not known to be normal. See
https://en.m.wikipedia.org/wiki/Five-number_summary
--Kip Murray
On Thursday, June 25, 2015, 'Bo Jacoby' via Programming <
[email protected]> wrote:
> NB. If you have got a bunch of numbers ]n=.?20#10035 32 66 37 94 56
> 44 30 25 70 27 68 49 92 96 79 74 99 72 22NB. and you want to simplify, then
> take the mean value
> 1 simplify n58.35
> NB. or take the mean value minus (or plus) the standard deviation 3": 2
> simplify n 33 84NB. or take three numbers having the same mean value and
> the same standard deviation and the same skewness as the original numbers
> 3": 3 simplify n
> 28 56 90NB. and so on:
> 3": 4 simplify n
> 26 46 68 94 3": 5 simplify n 25 37 60 75 96 1 simplify 5 simplify n
> 58.35NB. I think this simplify verb is cute and perhaps even useful, so I
> am sharing it with you.
> simplify=. 4 : 0
> y=.x*}.(%{.)+/y^/i.>:x
> x=.1
> for.y do.x=.((-/x*(#x){.y)%#x),x end.
> -|.>{:p.x
> )NB. If you can simplify the programming, the please let me know.
> ThanksBo
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
--
Sent from Gmail Mobile
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
