Article appearing in Shiken 19.1 (April 2015) pp. 32-41.
Authors: Trevor A. Holster1, J. W. Lake2
1. Fukuoka University
2. Fukuoka Jogakuin University
Abstract:
Smiley's experience reported in this issue of Shiken is probably quite typical of moving from traditional analysis to Rasch analysis. Traditional analysis, exemplified by Brown's (2005) Testing in Language Programs, provides statistics such as item facility values (IF) and item discrimination (ID) which will identify most of the same problematic items as Rasch analysis, and it's unlikely that classroom grades would change to any substantive degree between the two for a thoughtfully developed test. Rasch analysis provides benefits beyond analogues of traditional item analysis, however, and this paper argues that two important practical benefits are the variable map, or Wright map, which provides a quick visual summary comparing students with instructional features, and data-model fit statistics which provide for diagnosis and identification of students requiring remedial instruction. This study illustrates the potential of these for curriculum planning and classroom diagnosis through analysis of the vocabulary section of an academic English placement test.
Keywords: Diagnostic assessment, Rasch, item analysis, vocabulary testing