Here's a simple question, for which I was startled to find no easy and
definitive answer in the software manuals.
Given a contingency table with one dimension nominal (rows are groups)
but the other dimension ordinal (eg, a scale), what is the preferred or
most powerful test for a difference in the mean response between groups?
SPSS and Stata both offer chi-square as a general test of independence and
some derived measures of association, but offer alternatives tests only
when both variables are ordinal (eg, test of concordance). A Mann-Whitney
test based on ranks is an option, but with an ordinal scale I would expect
all the ties to render it somewhat pointless.
In standard form, might I suggest people send comments privately, and I will
summarise for the list? If I got enough replies, I could form a table:
V Good Good Poor Useless
Test 1 ....
Test 2 ... etc
R. Allan Reese Email: [log in to unmask]
Associate Manager Direct voice: +44 1482 466845
Graduate Research Institute Voice messages: +44 1482 466844
Hull University, Hull HU6 7RX, UK. Fax: +44 1482 466846
====================================================================
Hull: a 1st Division University (Daily Telegraph, London 20/8/97)
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|