On Tue, 14 Dec 1999 16:39:39 -0500 (Eastern Standard Time)
"Kenneth M. Steele" <[EMAIL PROTECTED]> wrote:
>
> On Tue, 14 Dec 1999 11:53:01 -0500 (EST) Stephen Black
> <[EMAIL PROTECTED]> wrote:
>
> > I've just been reading a new paper by Hodoba (Sleep Research Online,
> > 1999, at www.sro.org/1999/Hodoba/101 on the effect of chewing gum on
> > subjective feelings of sleepiness.
> >
> > He says yes, and among other data, reports a Mann-Whitney test (exptl
> > = 21, control = 43) with a median sleepiness score of 2 for the
> > experimentals (range 1-3) and median for the controls of 2 (range
> > 1-5). He reports a U of 282.0, Z of 2.6278, and significance at
> > p = 0.0086
> >
> > Question: is this possible?
> >
>
Both Mike Scoles and I suggested data sets that could produce
equivalent medians but different rank score sums on the M-W.
Both suggestions concerned odd distributions of scores.
If you go look at the paper cited by Stephen (the *.pdf
version) and examine the figures (which show boxplots for the
different groups) then you can see that the distributions are
very skewed or are otherwise not symmetrical.
Another notable aspect of the pdf version of the report is the
acknowledgment. Check it out.
Ken
----------------------
Kenneth M. Steele [EMAIL PROTECTED]
Associate Professor
Dept. of Psychology
Appalachian State University
Boone, NC 28608
USA