Exposure Control Procedures With the Generalized Partial Credit Model in Mixed-Format Tests

• In Event: Hilton Roundtable Session 4
  In Roundtable Session: Division D Section 1 Roundtable Topic: Applications of Item Response Theory

Abstract

This study compared two commonly used exposure control procedures, the randomesque and Sympson-Hetter methods, with the progressive-restricted standard error (PR-SE) method and no exposure control in a mixed-format testing context using the generalized partial credit model. The simulation study manipulated the test length and included the fixed- and variable-length stopping rules with a short and long test length. PR-SE administered nearly the entire test unit pool, where the minimum non-administered rate for any other procedure was 34%. PR-SE consistently resulted in the lowest over-exposure peak rates. Sympson-Hetter resulted in the smallest bias and root mean squared error but PR-SE showed no functional bias, and differences between methods were seen only in the extreme lower tail of the theta distribution.

Authors

• Audrey J. Leroux, Georgia State University

• Ian F. Hembry, MetaMetrics

• Barbara G. Dodd, The University of Texas at Austin