A Multimatrix Design and Multidimensional Item Response Theory Scaling of Math Skills: Links to Postsecondary Success in STEM

• In Event: Item Response Theory

Abstract

Students' high school-level math knowledge is a key predictor of later academic success in STEM fields. However, most diagnostic assessments model students' math competence as a unidimensional construct and provide limited evidence of its predictive validity for later success in STEM study programs. This paper thus describes the development of a math test with 11 subscales of high school-level math competency and tests its predictive validity for postsecondary performance in STEM (N = 3,217 STEM majors). A multi-matrix test design with 22 booklets was used, and multidimensional IRT analyses modeled the math subscales as separate, interrelated constructs. Different IRT modeling approaches were compared in terms of estimated test reliability and predictive validity in several student groups.

Authors

• Jeffrey M. DeVries, University of California - Irvine
  Presenting Author

• Daria Katharina Benden, University of Bonn
  Non-Presenting Author

• Fani Lauermann, University of Bonn
  Non-Presenting Author