|Title:||Bayesian Dynamic Borrowing: A Method for Utilizing Historical Data in Education Research|
|Principal Investigator:||Kaplan, David||Awardee:||University of Wisconsin, Madison|
|Program:||Statistical and Research Methodology in Education [Program Details]|
|Award Period:||3 years (07/01/19 – 06/30/22)||Award Amount:||$802,314|
|Type:||Methodological Innovation||Award Number:||R305D190053|
Co-Principal Investigator: Chen, Jianshen
In education research, large-scale educational assessments provide a wealth of historical data. The data from these assessments are typically collected via complex sampling designs, and the nuances of these designs must be accounted for. Bayesian dynamic borrowing allows a researcher to account for these nuances, including the fact that not all historical data, even from the same survey program, are of equal quality. Prior information can thus be systematically adjusted to reflect the analyst's degree of confidence in the importance and/or quality of sources of prior data. The purpose of this project is to develop and expand Bayesian dynamic borrowing to common education research designs applicable to large-scale assessments.
The research plan consists of three major parts. First, the research team will implement Bayesian dynamic borrowing under both the single-level and multilevel (i.e., nested) designs characteristic of large-scale assessment data. Second, the team will expand Bayesian dynamic borrowing to longitudinal designs focusing on growth over time. Third the researchers will expand Bayesian dynamic borrowing to problems of causal inference in observational settings using propensity score analysis. The research team will use a combination of Monte Carlo simulations guided by real data designs and applications of the developed methods to data from PISA, NAEP, and ECLS. Following any necessary changes to and retesting of the equations, the research team will develop software for conducting the multiplicity adjustments. They will also test the usability of the software and make changes as needed to ensure that it is user-friendly.
Products include a software package available as a Shiny app which researchers can use to systematically incorporate prior historical information into current Bayesian analyses. In addition to the software, the research team will disseminate results through peer-reviewed journal manuscripts, conference presentations, and workshops. The team will also create a webinar which will be available on MDRC's website and develop a user's guide for applied researchers. All disseminated products, including the Shiny app, will be accessible through http://www.bise.wceruw.org/.
Related IES Projects: Bayesian Inference for Experimental and Observational Studies in Education (R305D110001)