Influences of Carry-Over Effects Across Scales on Mediation Analyses

• In Event: Measurement and Statistical Modeling

Abstract

The one-statement-multiple-scale (OSMS) design often appears in psychological and educational surveys. However, the OSMS design may cause a non-recursive carry-over effect, because a preceding response of the same statement can affect its subsequent response. Consequently, the local item independence assumption of most measurement models is violated. Ignoring the carry-over effect may provide biased estimated item parameters for latter scales as well as the biased relationships between scales in an OSMS design. This study extended previous studies to investigate the consequences of model misspecification for OSMS items in mediation analyses via simulations and a real example from the Teaching and Learning International Survey (TALIS). Implications for dealing with the carry-over effects appropriately were also provided.

Authors

• Kuan-Yu Jin, Hong Kong Examinations and Assessment Authority

• Yi-jhen Wu, Technical University Dortmund