SRMS Webinar: A Comparative Review of Data Integration Methods

Abstract: With new challenges in data collection, a single data source often cannot meet the growing demand for high quality statistical products. Integrating multiple data sources has become a research priority, such as combining information from multiple surveys, combining surveys and administrative records, combining probability and non-probability surveys, and combining randomized experiments and observational studies. The key concern of data integration is to achieve transportability across different groups or data sources and population generalizability, which is the fundamental goal of survey inference. Hence, existing data integration methods are rooted in survey research and propose adaptations to overcome new challenges. In this webinar, we will review existing data integration methods, compare their assumptions and properties, and clarify their connections and variations with traditional probability survey inference approaches. The webinar aims to build up the fundamental concepts and quality criteria for an integrated data infrastructure