Re: st: Box and whiskers graph

[Marcello Pagano <Pagano@hsph.harvard.edu>]

What Ted said is quite correct, but you can also exercise your 
imagination and notice that if you have normal data, then the expected 
value of the interquartile range is 1.34 sigma (just reading from the 
normal tables).  Thus, in this case, the adjacent values are estimates 
of mean plus or minus twice the interquartile range, or mean plus or 
minus 2.64 sigma.  That means that you should have almost all your data 
(99.7%) between the two adjacent values.

m.p.

S�ren O'Neill wrote:


> Thank you very much for your reply - is there a reference you can give me, 
> that explains the reasoning behind these adjacent values - personally I rely 
> mostly on the book Practical Statistics for Medical Research by D. Altman and 
> I cant seem to find any reference to 'adjacent values' therein ... Also the 
> Stata Users guide and reference manuals seem to come up short ...
>
> Kind regards
> Soren
>
> On Monday 05 July 2004 23:32, Ted Anagnoson wrote:
>
>
> > The adjacent values separate the outliers from the rest of the data.  They
> > are the 25th or 75th percentiles plus 1.5 times the Inter-Quartlile range,
> > which is the distance between the 25th and 75th percentiles.  However, the
> > adjacent value indicated on the graph, by convention, is "rolled back" to
> > an actual data point, so that there is always real data underneath the
> > adjacent value.
> >
> > So the adjacent values are not actually percentiles....but "adjust" to the
> > characteristics of the data.
> >
> > Ted Anagnoson
> > California State University Los Angeles
> >
> > At 10:46 PM 7/5/04 +0200, you wrote:
> >
> > > The box and whiskers graph illustrates the median value (with a line), the
> > > 25%
> > > and 75% centiles (with a box) and the adjacent values (with the whiskers)
> > > ... but what are the adjacent values? 2� and 97� centiles or ... ???
> > > --
> > > S�ren O'Neill, kiropraktor
> > > soren.oneill@mail.dk
> > > tel. 6362 1906

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/