For a psychometric survey of 50-60 items using a
9-point Likert-type response scale, with mildly
complex instructions for choosing the extreme
anchors, subjects should be dropped if they choose the
same anchor-point for more than 35 items. E.g., if a
subject selects four 36 times or more, that subject
is eliminated from the analysis.
A colleague administered the survey, in Chinese,
online in China using a paid online research
panel. Approximately 33% of the subjects would be
eliminated if the 36-or-more identical response rule
is
applied.
From my research using the same instrument in
China I found it necessary to drop less than 5% of the
subjects using hand delivered and picked up paper
forms, and from a report by the author of the
survey, his experience is less than 10%.
From my meagre experience of living and working in
China for several years and engaging in survey
research there for 12 years, I would speculate
that in the minds of the subjects there is no
relationship existing between the members of the paid
online research panel and those collecting the data,
other than being paid more for completing more
surveys,
leading to minimising the amount of time spent on
each survey, by, e.g., the �malicious response set
behaviour� of using as few anchor points as
possible.
I have not yet done a literature search.
Your thoughts? Or knowledge of the literature? Or
research results?
"Who dare to teach must never cease to learn."-John Cotton Dana
Romie F. Littrell, BA, MBA,PhD, FAIR, An f�na� fi�in
Faculty of Business, Auckland University of Technology, N.Z.
http://www.romielittrellpubs.homestead.com/
http://www.crossculturalcentre.homestead.com/
PARTICIPATE in a study of leadership & values:
hppt://www.leadershipvalues.homestead.com/
___________________________________________________________
Yahoo! Messenger - NEW crystal clear PC to PC calling worldwide with voicemail http://uk.messenger.yahoo.com