Hmmm... If MaxDiff is what you have found, then I'm surprised that the OP
didn't know about it. Sounds like a common (but perhaps new) method in market
research, which he says is his field. Curious situation...


On Mon, 3 Oct 2005 21:16:01 -0700, "
wrote:

"market researcher needs help!" wrote:
A client at work asked us to show (a) MaxDiff with our survey results (we
have written 3 surveys for them). I havent been able to get in touch with
them, and one of my coworkers thought it was an excel function, so Im
trying
to find out what they meant by it. Any insight you can offer would be
greatly appreciated!

You really should learn how to use google search. I suspect this is what
your client is asking about. Quoting from one google "hit" ....

"Maximum Difference Scaling is ideally suited for this task

Instead of asking respondents to directly rate or rank their preferences
for these flavors, we used a MaxDiff approach*

Each respondent was shown 10 sets of four flavors

For each set, we asked them to pick the flavor they like best of the four,
and the flavor they like least

From these responses, we can derive their preferences for all 12 of the
flavors

* Traditional rating scale and ranking questions often fail to produce
useful results. Problems include halo effects, skewed distributions (limited
usage of the full scale range), many ties among items, and response style
biases (some respondents refuse to give top box ratings, while others only
use the top 3 boxes). Maximum Difference Scaling is a relatively new
approach that overcomes these pitfalls, producing true derived
interval-scaled data."

For a more formal explanation, see
http://www.sdr-consulting.com/article19.html .