Randomized Trials and Observational Data: Harnessing the Best of Two Worlds
One type of experimental data, the randomized controlled trial (RCT) data is often regarded as the “gold standard” of causal inference in health sciences, and are increasingly common in other fields. They allow for an unbiased estimate of an intervention’s causal impact. However, RCT sample sizes are often small and the results can be too imprecise to guide decision making. Observational studies, by contrast, offer much larger sample sizes, but may suffer confounding.
In many cases, experimental and observational data exist side by side (such as patient record data vs. trial data; student record data vs. intervention data; convenience sample vs. representative sample). This working group focuses on methodological issues for integrating “big observational data” with “small but high-quality experimental data” to get the best of both worlds. Attendees will discuss issues such as data access, data linkage, transfer learning, treating subgroups, improving trial design and external validity.
We welcome all faculty members interested in these topics. The focus of the discussion will be organically decided by the participants. Through these discussions, we hope to foster ideas and collaboration that will lead to improved projects and new grants.
Working Group leader: Johann Gagnon-Bartsch, Assistant Professor, Statistics
Day 1 (10/2/2020, 11:00am):
- Johann Gagnon-Bartsch, Assistant Professor, Statistics: Combining observational data and RCT – an overview
- Xu Shi, Assistant Professor, Biostatistics: Negative controls for observational data
Day 2 (10/7/2020, 3:00pm):
- Christopher Gillies, Assistant Research Scientist, Emergency Medicine: Clinical Early Warning Systems: Transfer Learning and Validation using RCT
- Stephan Taylor, Professor and Associate Chair for Research and Research Regulatory Affairs, Psychiatry, and Daniel Almirall, Research Associate Professor, Survey Research Center: RCT and medical decision-making
Day 3 (10/9/2020, 11:00am):
- Christina Weiland, Associate Professor – School of Education: RCT and observational data in early childhood education research