Monitoring trends in education outcomes over time is critically important to education policy. Dr. David Kaplan, the Patricia Busk Professor of Quantitative Methods,
and his graduate student Mingya Huang have developed a Bayesian probabilistic forecasting workflow that can improve this monitoring when applied to large-scale assessment trend data. At the international level, large-scale assessment programs such as the OECD-sponsored Program for International Student Assessment (PISA) (OECD, 2001) can provide information to be used to forecast movement toward the goals set by the United Nations. One such goal, for example, is achieving literacy and numeracy for men and women (Goal 4.6 of the United Nations Sustainable Development Goals). In the U.S., the National Assessment of Educational Progress (NAEP) has provided long-term trend data for the nation since 1970 and for each state since 1996. Using methods borrowed from economic forecasting and weather forecasting, Kaplan and Huang found that the Bayesian perspective of accounting for all forms of uncertainty in the modeling process can lead to a richer description of national trend data than what is typically reported in official documents. Kaplan will be continuing this work using PISA data in collaboration with colleagues at the University of Heidelberg.
The full paper, published in Large-Scale Assessments in Education, can be found here.