Tuesday, April 12, 2022
2:00 - 4:00 PM (PDT) on Zoom
Researchers interested in participating in practice sessions should email convenors@berkeley.edu.
Talk 1:
Xingyao (Doria) Xiao: Bayesian Growth Mixture Models for Classifying and Measuring Individual Trajectories
Abstract:
Growth-Curve Modeling (GCM) is a tool for measuring interindividual differences in intraindividual change on a continuum. Growth Mixture Modeling (GMM) adds the capability to examine heterogeneity in the between-individual trajectories due to unobserved classifications of individuals. We introduce GMMs and demonstrate Bayesian estimation in the Stan package.
Talk 2:
Kristin Gunckel and Linda Morell: Examining the crosscutting concept of patterns: An initial construct map in the context of ecosystems
Abstract:
We present work to assess student competencies for the crosscutting concept of patterns. Our research is based on the BEAR Assessment System, grounded in four building blocks (e.g., construct map, item design, outcome space, and measurement model) which guide assessment development. Our first challenge was to develop a suitable construct map. For this, we focused on studentsâ increasingly sophisticated thinking about patterns in ecosystems. We then developed assessment items that included representations of these patterns using both graphs and maps. The items were administered to 176 middle school students and scored using the initial levels in the construct map. Findings indicate that items that prompt students to describe specific patterns were qualitatively easier for students than items that asked students to describe patterns they might notice. Students also experienced more difficulty noticing and describing static patterns. Familiarity with representations and relevant disciplinary knowledge may also influence student proficiency in finding patterns. These findings will inform the eventual development of a learning progression for the patterns crosscutting concept. Given the fieldâs limited understanding of the crosscutting concepts, let alone how to assess them, we envisage this work will be of considerable interest to NARST members.
Talk 3:
Perman Gochyyev: Modeling a summative assessment for a multidimensional learning progression
Abstract:
Learning progressions (LPs) are often multidimensional due to the scope of the learning represented. We investigated qualities of different approaches to estimate student ability at the overall macro level, in addition to each underlying dimension (or construct), in the context of a college-readiness assessment in problem solving using mathematics (PSM).
Talk 4:
Joshua Sussman: Examining Differential Rates of Progress Along an Early Childhood Learning Progression
Abstract:
We describe findings from a new psychometric method related to differential item functioning designed for use with longitudinal data collected from the perspective of a learning progression. We modeled the emergence of learning differences between groups of children and identified specific learning subdomains that contributed most to the observed differences.