Practical Improvements in Causally-interpretable Meta-analysis and Transportability

Presented by Kollin Rott
Ph.D. Candidate in Biostatistics
Ph.D. Advisers: James Hodges & Jared Huling

This dissertation has a clear overarching theme: improving and spreading the practical application of causal transportability methods. The first two papers focused on causally-interpretable meta-analysis. To avoid overwhelming the audience, I present only the third paper today. We propose a new approach to balancing weights, in which the degree of balance chosen for a given covariate depends on its strength as an effect modifier. We view this approach as a promising development in weighting, as it is more stable than inverse probability weighting techniques and fails to admit a solution much less often than prominent balancing weights techniques. The framework is flexible, heightening its practical appeal and opening many directions for further research.