An Application of Multiple Correspondence Analysis to the Experimental Study of Item Context

• In Event: Psychometrics and Educational Measurement Poster Session 2

Abstract

Though item contexts are common features in assessments that range from the classroom to international level, their effect on student performance is not well understood. Additionally, contexts themselves are a complex system of content, linguistic factors, illustrations, and cultural references that make their systematic study difficult. In this paper, we describe an approach for the principled study and manipulation of item contexts using the data reduction technique of multiple correspondence analysis. We apply this technique to a middle school physics test and experimentally test the effect of context manipulations on a sample of N=3400 respondents. We find the approach key for minimizing researcher degrees of freedom and identifying focused manipulations to contexts for experimental applications.

Authors

• Klint Kanopka, Stanford University

• Philip Hernandez, Stanford University

• Nixi Wang, University of Washington - Seattle

• Maria Araceli Ruiz-Primo, Stanford University

• Min Li, University of Washington

• Jim Minstrell, FACET Innovations