Date: Tue, 5 May 1998 14:32:50 -0500
Reply-To: "Nichols, David" <nichols@SPSS.COM>
Sender: "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From: "Nichols, David" <nichols@SPSS.COM>
Subject: Re: mean rankings in nonparametric tests
If those two individuals had positive values on the dependent, while others
had negative values, then they would have the highest ranks. Data are
ranked from low to high on the dependent variable for the Mann-Whitney
test.
David Nichols
Principal Support Statistician and
Manager of Statistical Support
SPSS Inc.
nichols@spss.com
----------
From: Earthstars [SMTP:Earthstars@AOL.COM]
Sent: Thursday, April 30, 1998 3:50 PM
To: SPSSX-L@UGA.CC.UGA.EDU
Subject: mean rankings in nonparametric tests
I have a similar problem related to one recently posted about
Mann-Whitneys
and rankings.
My highest mean rank should have logically been the lowest ranking (based
on
the hypothesis and in conjunction with numerous other significant results
in
the correct direction); e.g., The higher the degree of infiltration of
managed
care at educational institutions, the more the director's morale and
quality
of life improved (inconsistent with every other finding, including
increased
clinical responsibilities of directors in institutions with increased
managed
care infiltration).
There were only 2 individuals who indicated that their morale and quality
of
life improved with managed care and 48 individuals who said it decreased
with
higher degrees of managed care infiltration. Would this have effected the
results? (How could the ratings of those 2 individuals outrank the 48?) If
it
does effect the results, how should they be interpreted and how many
subjects
per group are needed for valid conclusions to be reached? Is there a
correction for unequal groups?
Additionally, how should it be interpreted when Mann-Whitney does not
correct
for ties? Do the ties supposedly cancel each other out? Thanks.
Cindy
