Evaluating Dichotomous and Partial-Credit Scoring Within a Constructed-Response Assessment: Is More Information Always Psychometrically Better?

• In Event: Hilton Roundtable Session 1
  In Roundtable Session: Development, Validation, and Psychometric Testing of Assessments

Abstract

While literature on dichotomous versus partial-credit scoring models is plentiful, research comparing the two models across the same items is sparse, focusing primarily on multiple-choice tests. This ignores an important type of item used with particular frequency in mathematics. The present study compared two models of scoring constructed-response items across aspects of psychometrics, impact, and validity. Scores from a mathematics problem-solving test for sixth-grade (PSM6) were calculated using both models, explored, and examined. While educators may feel uncomfortable using a dichotomous scoring model to evaluate constructed-response items, our study demonstrated near equivalence across psychometric properties, student ability outcomes, and in relation to other variables validity. Such equivalence supports the optional use of a dichotomous model for assessing constructed-response items.

Authors

• Toni A. May (Sondergeld), Drexel University

• Gregory E. Stone, University of Toledo

• Lance Kruse, North Carolina State University

• Jonathan David Bostic, Bowling Green State University

• Gabriel Matney, Bowling Green State University