Abstract
The COVID-19 pandemic forced millions of students to transition from traditional in-person instruction into a learning environment that incorporates facets of social distancing and online education (National Center for Education Statistics, 2022). One consequence is that the massive disruption of the COVID-19 health crisis is related to the largest declines in elementary and secondary students’ educational achievement as inferred from recent results of the National Assessment of Educational Progress long-term trend (U.S. Department of Education, 2022). Accordingly, recent events have raised awareness of the need for robust formative assessments to accelerate learning and improve educational and behavioral outcomes.
The availability of the existing data sets and designs, online learning technology, handheld devices, and wearables provide researchers with new opportunities for designing diagnostic methods or applying the existing statistical and psychometric tools for improving educational outcomes. To meet the scientific challenges associated with enhancing student learning, additional methodological and statistical research are needed to advance and refine methods for identifying novel interventions. The Journal of Educational and Behavioral Statistics has served as a long-time outlet for disseminating methods and for providing a forum for discussion and debate on the role of statistical methods for the educational, social, and behavioral sciences. The purpose of this special issue is to highlight statistical methods for providing decision-makers and users with fine-grained information to improve educational and behavioral outcomes.
This special issue includes papers in two general areas of statistical methodology and application. The first collection of papers focuses on new methods and novel applications of diagnostic models within the framework of restricted latent class models. Liang et al. (2023) present new methods for incorporating covariates into longitudinal diagnostic models that aim to track and monitor learning using a computationally efficient three-step estimation method. Nájera et al. (2023) advance parsimonious models for leveraging classroom-level data to provide educators with reliable diagnostic information about student skill mastery. Su and Henson (2023) demonstrate practical issues for implementing and designing diagnostic assessments systems and report results from a novel application. Yamaguchi (2023) proposes a flexible model for uncovering relationships between student skill profiles and contextual profiles derived from explanatory variables, such as student background questionnaire items, using classical and variational Bayesian methods. Chen and Wang (2023) introduce methods for inferring the underlying graphical structure of student skills to infer the nature of the hierarchical structure of skill mastery, which is a central concern for designing learning interventions. Li et al. (2023) demonstrate the utility of deploying diagnostic models in computerized adaptive testing (CAT) in an application involving elementary students Chinese reading comprehension. The second collection of papers are grounded in the factor analytic tradition and provide novel extensions and applications to identify and detect beneficial educational interventions. Yu and Douglas (2023) extend the classical item response theory (IRT) model by incorporating item-specific learning parameters and proposing methods for adaptively selecting items and interventions to accelerate learning. Understanding factors that inform heterogeneity of treatment effects is a central concern for the educational sciences, and Gilbert et al. (2023) propose new methods for modeling treatment effect heterogeneity within the IRT framework at the item level. Tree models provide a powerful framework for broadening the utility of IRT models for novel educational and behavioral assessments. Davison et al. (2023) introduce a novel tree model and application focused on an inferential reading comprehension system using a sequential, multidimensional IRT measurement model with CAT. We hope this special issue sparks continued interest in the role of statistical methodology in supporting the use of data to inform timely, diagnostic educational decisions.