2026-27 Seminars and Recordings
September 25 , 2026, 11 am CST
MetaMate: Automating Study Coding for Educational Systematic Reviews with Large Language Models
-
Speaker: Dr. Xue Wang, Johns Hopkins
-
Description: Data extraction is the slowest, most error-prone stage of a systematic review. MetaMate is a free, open-source web tool that uses large language models to extract study-level data from uploaded PDFs in seconds, using a customizable coding scheme and multimodal parsing of tables and figures. In evaluations against trained human coders on two datasets (32 and 113 studies), MetaMate matched or exceeded coder accuracy on most participant and intervention elements. This session demonstrates MetaMate live on real education studies, shares what we learned about where it succeeds and fails, and offers practical guidance for using AI extraction responsibly in your own reviews.
-
Video Recording
October 30th, 11 am CST
Selecting Meta-Regressors in Meta-Analysis: A Principal Component Analysis Approach
-
Speakers: Dr. Bixi Zhang, CUNY Graduate Center, & Dr. Jay Verkuilen, CUNY Graduate Center
-
Description: Meta-regression helps explain effect size heterogeneity, yet numerous, correlated, or incomplete moderators complicate model specification. This study examines weighted principal component analysis as a preprocessing approach for selecting metaregressors while incorporating random effects weights. A Monte Carlo simulation varied key design parameters and compared weighted with unweighted PCA under a multilevel data structure. Results indicate that the proposed approach can recover the underlying structure with moderate numbers of studies, with modest benefits from weighting. Applied examples illustrate how dimension plots can summarize moderator structure, visualize relationships among variables, and support interpretation before metaregression.
-
Video Recording