Multilevel Multiple-Group Latent Profile Analysis of Science Instruction: Cross-National Comparison of Korea and Japan

• In Event: 35.073 - Innovations in Survey Research Methodology

Abstract

This study proposed a new approach for Multilevel Multiple-group Latent Profile Analysis (MMLPA) with an illustration. MMLPA is advantageous since it compares the characteristics of latent profiles across countries after the similarity of latent profiles is explicitly examined and guaranteed, considering the dependency among observations. This study analyzed instructional practices in science classrooms in Korea and Japan using MMLPA. The result showed that a multilevel LPA was more proper than a single-level LPA for the nested data. In the next steps, the similarity of multilevel latent profiles was satisfied. After establishing the similarity of latent profiles, dispersion, distributional, predictor, and outcome similarity were examined to decide whether the differences of latent profiles could be interpreted in the context of countries.

Authors

• Hyun Jeong Park, Seoul National University

• Yoonhee Son, Korea Institute for Curriculum and Evaluation

• Yujung Hong

• Junok Kim, University of California - Los Angeles