This is a commentary on “Updating the Duplex Design for Test-Based Accountability in the Twenty-First Century,” by Isaac I. Bejar and E. Aurora Graf.
Notes
1After recognizing that a partial rotation of the item clusters across the possible four positions in a booklet leads to a bias in the item difficulty estimates, the item clusters were rotated across all four possible positions in the 2003 study (CitationOECD, 2002, Citation2005).
2Due to large position effects for items in the third (the last before the break) and the sixth (the last before the end) cluster in the booklets of the 2003 study, it was decided to change to a design with four clusters per booklet for the 2007 study (CitationMartin, Mullis, & Chrostowski, 2004; CitationOlsen, Martin, & Mullis, 2008).