Applying Partial Credit Trees as a Multi-Indicator DIF Screening Tool in Scale Development

• In Event: AERA Poster Session 8
  In Poster Session: Test Validity Research and Evaluation (San Diego)

Abstract

The purpose of this study is to use partial credit trees, a relatively new differential item functioning (DIF) procedure that allows for multiple group comparisons within a single analysis, as a pilot screener for item performance in new scale development. An abortion legality attitude scale with four response formats was investigated: four-, five-, six-, and seven-points. The four item sets were analyzed for DIF that might be influenced by response format as these same items exhibited DIF with a two-point response condition. Results indicate no difference in threshold parameters for subgroups of respondents as identified by four demographic variables for all four scales which contrasts with the prior results from using a dichotomous version of the same scale.

Authors

• Nana Amma Asamoah, University of Arkansas

• Ronna C. Turner, University of Arkansas

• Wen-Juo Lo, University of Arkansas

• Brandon Crawford, Indiana University - Bloomington

• Kristen Jozkowski, Indiana University

• Sara McClelland, University of Michigan