A New Approach to Modeling Directional Local Item Dependence in Multipart Items Evaluated With Conditional Multivariate Distribution

• In Event: 34.079 - Roundtable Session 14
  In Roundtable Session: 34.079-11 - Ordinal and Binary Data: From Differential Item Functioning to Classification and Regression Tree

Abstract

Local item dependence (LID) means the items are clustered around a common stimulus. The common stimulus is often called a testlet (Wainer & Kiely, 1987). In testlets, local item dependence is most often non-directional. LID could be directional as well.
This study proposes a new approach to accounting for directional local dependence in multipart items. The new method models directional local item dependence between two items by estimating the correlation of difficulty parameters of two items that are dependent by assuming the bivariate normal relationship among the difficulty parameters. Markov Chain Monte Carlo method is used for model parameter estimation. Model parameter recovery is evaluated in a simulation study in terms of item and ability parameter estimation.

Authors

• Kaiwen Man, University of Alabama

• Hong Jiao, University of Maryland-College Park

• Yunbo Ouyang, University of Illinois at Urbana - Champaign

• Yong Luo, National Center for Assessment, Saudi Arabia