A Reliability of Separation Index for Evaluating Unfolding Scales (Poster 5)

• In Event: AERA Virtual Poster Session 8
  In Poster Session: Poster Session 1

Abstract

Unfolding models are widely used in attitude assessments, and provide more information than the dominance model. The dominance models produce monotonically increasing probability functions, and the unfolding models generate single-peaked probability response functions. On an unfolding scale, each respondent has an ideal-point location indicating their preference or attitude on the latent continuum. The relative distance between a person's ideal-point and an item's location indicates the probability an individual will endorse an item. Classical test theory based reliability indices (e.g., coefficinet alpha) are inappropriate for evaluating unfolding scales for a variety of reasons. Empirical and simulation studies are conducted to demonstrate the uses of this reliability index for unfolding models.

Authors

• Trenton Combs, University of Miami

• Jue Wang, University of Science and Technology of China