PISA Test Anxiety Predictor Exploration via Penalized Regression (Poster 7)

• In Event: AERA Virtual Poster Session 7
  In Poster Session: Poster Session

Abstract

An emphasis of large-scale international studies is the collection of diverse data to inform policy and practice to improve student learning. To-date, methodological challenges have restricted the utility of traditional regression approaches to identify predictors of student outcomes. Recent developments in machine learning have advanced penalized regression approaches capable of handling large number of predictors for variable selection for model generation. In response, the penalized regression procedures of least absolute selection and shrinkage operator and elastic net were used to analyze Programme for International Student Assessment (PISA) 2015 data to identify predictors of U.S. students’ (N = 5,593) test anxiety. Of 182 predictors, 27 were identified as key indicators of test anxiety. Implications to research and practice are discussed.

Authors

• Jin Eun Yoo, Korea National University of Education

• Tai Sun Jeong, Korea Institute for Curriculum and Evaluation

• Jason C. Immekus, University of Louisville