Article appearing in Shiken 26.1 (June 2022) pp. 20-37; Issue DOI: https://doi.org/10.37546/JALTSIG.TEVAL26.1
By Jordan Svien
Hiroshima Bunkyo University
Article DOI: https://doi.org/10.37546/JALTSIG.TEVAL26.1-2
Abstract:
From 2015 to 2019, the Bunkyo English Communication Center at Hiroshima Bunkyo University conducted end-of-semester speaking exams called Bunkyo English Speaking Tests (BESTs) for all English Communication freshman and sophomore students. During these five years, the Bunkyo English Communication Center learned several test administration best practices. First, in a desire to apply a many-facet Rasch model using the Facets software package (Linacre, 2022a) to provide student fair scores that account for rater leniency and severity, a preventative flaw in the rater schedule was discovered and corrected. Second, the increased complexity of the rater schedule plus a desire to streamline the exam processes necessitated the building of a comprehensive scheduling and testing system in Excel. Finally, the calculation method initially used for converting Rasch measures into student fair scores was based on a faulty assumption and suffered from ambiguity and subjectivity, and a fairer workaround system was discovered and implemented. This paper documents the discovery of these problems and the process of developing and implementing their solutions.
Keywords: examination, assessment, MFRM, Facets