Using Latent Class Analysis to Validate a Classroom-Based Numerical Reasoning Tool

• In Event: Hilton Roundtable 2
  In Roundtable Session: Using Data as Evidence for Improving Student Outcomes

Abstract

Using Cronbach and Meehl’s (1955) approach to establishing construct validity, operational definitions are proposed for various types of numerical reasoning strategies on simple whole number computation and fraction comparison items. Definitions were based on the collection and analysis of over 5,000 responses to
a minimum of 10 items completed by 523 intermediate-aged students. Definitions were used to create a formative assessment framework validated through use of cluster analyses, hierarchical linear modeling, a one-way ANOVA, and a latent class analysis. The five categories proposed include: (1) Absence of reasoning, (2) Faulty reasoning strategies, (3) Partial reasoning strategies, (4) Algorithmic reasoning strategies, and (5) Conceptual reasoning strategies. Examples are provided related to use of this framework in the classroom.

Authors

• Lindy L. Crawford, Texas Christian University

• Jacqueline Huscroft-D'Angelo, University of Nebraska - Lincoln

• Sarah Quebec Fuentes, Texas Christian University

• Matthew Lambert, University of Nebraska - Lincoln

• Kristina Higgins, Tarleton State University