Abstract:
|
Bayesian hierarchical models produce shrinkage estimators that can be used to integrate supplementary data into the analysis of a primary data source. Existing methods may be limited since they require either pre-specification of a shrinkage weight for each source or rely a single parameter for smoothing which risks considerable bias or minimal borrowing. We introduce multi-source exchangeability models (MEMs), a general Bayesian approach for integrating multiple, potentially non-exchangeable, supplemental data sources. MEMs induce source-specific smoothing parameters that can be estimated in the presence of the data to facilitate a dynamic multi-resolution smoothed estimator that is asymptotically consistent. When compared to other Bayesian hierarchical modeling strategies, we show that MEMs may achieve up to a 115% increase in effective supplemental sample size with exchangeable supplemental sources, as well as a 56% reduction in bias with heterogeneous supplemental sources. We illustrate the application of MEMs using a recently completed randomized trial of very low nicotine content cigarettes, which resulted in a 30% improvement in efficiency compared to the standard analysis.
|