Using the Rasch model to determine form equivalence in the trilingual Lollipop Readiness Test.

Determining the cross-cultural equivalence of multilingual tests is a challenge that is more complex than simple horizontal equating of test forms. This study examines the functioning of a trilingual test of preschool readiness to determine the equivalence. Different forms of the test have previously been examined using classical statistical techniques on bilingual forms (Lang, Chew, & Schomber, 1992). A family of psychometric models that maximize information from the data while independent of incidental circumstances is the Rasch model (Wright & Mok, 2004). The Rasch model met the requirements of analysis appropriate with different aggregations of items, handling of missing data, and sensitivity to misfit from the construct. Graphic and parametric statistical analyses resulted from the Rasch analysis that identified test items that functioned differently by language. Even though conclusions based on this sample size and the limitation of analysis with no common people or items would require further investigation after some test corrections, this exercise was revealing to the authors and convincing that the French version of the Lollipop Test was not performing as equivalent compared to the Spanish and English versions. This was completely unknown prior to this study.