Online Research Workshop
Aggregating Expert Ratings with IRT Models: The V-Dem Methodology
When: February 10, 2022 (Thursday) 5 pm to 6:30 pm (Japan Standard Time)
This workshop provides a conceptual and technical introduction to V-Dem’s methodology for aggregating expert ratings in cross-national research. The workshop first presents the reliability and validity challenges associated with the task and then introduces the components of the methodology designed to handle them. The methodology combines survey design elements, chiefly anchoring items, with model-based aggregation procedure that rests on a custom-developed Bayesian Item Response model implemented in the Stan probabilistic programming language. The workshop concludes with a technical discussion of the model and ways in which it can be adapted to other similar tasks.
Juraj Medzihorsky is an Assistant Professor in Social Data Science at Durham University with an appointment at the School of Government and International Affairs and the Research Methods Centre, and Program Manager for Statistical Computing in the V-Dem project. Previously he held positions at LSE, the University of Gothenburg, and CEU, where he also earned his PhD. His methodological interests center around measurement with surveys, text, and votes, and on causal inference in comparative research. His substantive interests focus on elections and processes of democratization and autocratization. His research appeared in journals including Political Analysis, PS: Political Science and Politics, and Research & Politics.