Application of Propensity Score Analysis in Linking Without Anchor Items

• In Event: 68.062 - Poster Session 15
  In Poster Session: 68.062-4 - Poster Session: Innovative Research on Diverse Methodological Approaches

Abstract

This study focuses on a test linking problem when both anchor items and equivalent groups are unavailable. Propensity score analysis is used to match examinees based on background variables, so as to create roughly equivalent groups for linking. A simulation study is conducted to evaluate the effectiveness of using propensity score matching in linking under two scenarios. In scenario 1, all background variables related to group equivalence are available; in scenario 2, one relevant variable is missing. Different caliper sizes are also considered to control the comparability between matched groups. Preliminary results in scenario 1 showed that using propensity score matching greatly reduced systematic linking error compared to using unmatched groups.

Authors

• Xi Wang, Measured Progress

• Hwanggyu Lim, University of Massachusetts

• Fei Chen, University of North Carolina - Chapel Hill