Paired Comparisons

Paired Comparisons - Tests

Thread view:

Enable EMail Alerts

Start New Thread

Thread rating:

KevinUK - 27 Sep 2009 12:45 GMT

I'm setting up a psycholinguistic experiment using the method of
paired comparisons. Participants will choose between two words/
entities, or signal 'no difference', based on a specific attribute.
The design is OK (I think) but I am struggling (as ever) to figure out
the appropriate statistical analysis. Possibilities encountered in
long Google searches include Bradley-Terry, Thurstone Case V, and
Vector Preference, all of which are currently a mystery to me!

The least I want is a ranking of the entities according to the
attribute,with appropriate measures of significance and power (each
participant would compare 15 entities, a total of 105 comparisons). If
I could also measure the 'distance' between each entity on a ranked
scale, then that would probably reduce the need for further
experiments.

So, I am looking for any or all of the following:
- A report of a similar experiment (in any field) that I might use as
a model
- Bradley-Terry, etc For Dummies (a cookbook explanation)
- Any software that might help
Thanks
Keith Aaron

Reply to this Message

Art Kendall - 27 Sep 2009 14:45 GMT

Please provide more detail about your situation. What questions are you
using the data to answer?

How are you going to present the stimulus pairs? computer screen, paper
form, over the phone, interview?

> If I could also measure the 'distance' between each entity on a ranked
> scale, then that would probably reduce the need for further experiments.

Do you mean distance on a* rating *scale?e.g., Little or no X, less
than average X, average X, more than average X, (almost) completely X
Do you want to have two stimulus by stimulus matrices of responses, one
for "dominance" or preference for row or columns stimulus, and another
for e.g.,similarity?

Or do you want to have two stimulus by stimulus matrices of responses,
one for "dominance" or preference for row vs columns stimulus, and the
other how much the subject prefers the chosen stimulus?

Will subjects judge the stimuli on different attributes so that you have
separate matrices for each attribute? What "attributes"?

Art Kendall
Social Research Consultants

> I'm setting up a psycholinguistic experiment using the method of
> paired comparisons. Participants will choose between two words/
[quoted text clipped - 19 lines]
> Keith Aaron
>

Reply to this Message

Math Forum / Math Software / SPSS / September 2009

Paired Comparisons - Tests