Guille RA. — American Board of Internal Medicine
Presented: American Educational Research Association Conference, April 2009
Abstract: This study examined how ancillary information collected when using the Weighted Deviations Model (Swanson & Stocking, 1993) for assembly of linear test forms could identify shortages in the pool from which the items were drawn. The study showed how, in addition to identifying constraint lower bounds not met, a light “stress test” of the pool could be performed immediately following form selection. The test yields an item statistic, referred to as “delta deviation,” which quantifies the relative scarcity of the selected items with respect to the encompassing item pool. Results showed that the delta deviation statistic identified items in historically scarce areas that were not revealed by simply examining the lower-bounds-not-met. The items could serve as exemplars to be modeled.
For more information about this presentation, please contact Research@abim.org.