|
Abstract:
|
The Weather Risk Attribution Forecast (WRAF) is a real-time forecasting tool that uses global climate models to examine whether and how anthropogenic emissions have contributed to extreme weather across the globe. The forecast involves making a collection of statements regarding the magnitude of this contribution; however, in conducting a large number of simultaneous hypothesis tests, the WRAF is prone to identifying false "discoveries." False discovery rate (FDR) control is a common approach for addressing this multiple testing problem, but unfortunately generic FDR procedures suffer from low power under dependence, and more refined techniques that account for the dependence are sensitive to model misspecification. Here, we develop a Bayesian decision theoretic approach for dependent multiple testing that flexibly controls the FDR and is robust to model misspecification. We illustrate the robustness of our procedure to statistical model error with a simulation study, using a framework that accounts for generic spatial dependence and allows the practitioner to flexibly specify the loss function criteria. Finally, we apply the procedure to several seasonal forecasts.
|
