Funded by:

A Bayesian model framework for analyzing data from longitudinal large-scale assessments

PIs:
Dr. Christian Aßmann, Leibniz Institute for Educational Trajectories
Prof. Dr. Claus Carstensen, University of Bamberg
Prof. Dr. Steffi Pohl, Freie Universität Berlin

Staff:
Luise Fischer, University of Bamberg
Theresa Rohm, University of Bamberg
Thorsten Schnapp, University of Bamberg
Daniel Schulze, Freie Universität Berlin

Project Summary:
The aim of our project is to develop a model framework for the analyses of data from longitudinal large-scale assessments (LSAs) addressing the different challenges the data pose. In the previous funding phase, we developed a Bayesian estimation approach, in which competencies are characterized as latent variables while simultaneously accounting for missing values in background variables. The approach is able to account for longitudinal data structures and also incorporates hierarchical structures.
In the current funding phase, we will extend this approach by further features. First we will implement statistical approaches for the comparison and averaging of non-nested model specifications and will make use of Bayesian model averaging for linking purposes (Aßmann & Pohl). Second, the model framework is extended to deal with log data. Specifically, we will incorporate recent innovations in response time modeling allowing for a more efficient competence estimation, as well as for modeling test taking behavior (Pohl & Aßmann). Third, we will extend our framework with regard to selection of background variables in a longitudinal setting for the estimation of plausible values. We plan to incorporate automated variable selection procedures to deal with the huge number of variables (Carstensen & Aßmann). We will demonstrate the use of our framework in applications using NEPS data. By further enhancing our current framework to incorporate model comparison and averaging, new linking approaches, the inclusion of response time models, and automated variable selection for estimating plausible values, our framework is applicable to a wider range of research questions typically addressed in LSAs. The estimation routines will be available to data users in form of R packages.