[Rasch] Best-Worst Average or Rasch?

Donald Bacon dbacon at du.edu
Wed Jul 27 02:51:53 EST 2011

Hi Rasch Experts –

A colleague of mine has a dataset generated from a best-worst questionnaire.  Respondents were given sets of three objects and asked to identify the most preferred and least preferred in each set.  An object might score 1 if most, -1 if least, and 0 if neither.  The sets are nicely rotated and balanced (9 objects, 3 at a time, 12 iterations), so taking the average score for objects across sets should be a reasonable metric of preference.

Obviously, the same data could be scaled more elegantly with the Rasch model.  What are the advantages of the Rasch model in this context?  Would it handle floor and ceiling effects (all 1’s, all -1’s) better or worse?  Would Rasch measures be so highly correlated with simple averages that I should not expect to see any differences in my subsequent statistical tests?  Are there some good cites on this issue that I should be reading?

In advance, thanks --

Don Bacon

Professor of Marketing
Incoming Editor, Journal of Marketing Education
dbacon at du.edu<mailto:dbacon at du.edu>, (303) 871-2707
Daniels College of Business
University of Denver


More information about the Rasch mailing list