Robust Decisions From Modeled Estimators

Tyler McCormick , University of Washington, Seattle

Despite recent advances in methods and technology, collecting data on population health and inequality remains financially and logistically taxing, particularly in low and lower-middle income countries. In such settings, policymakers in areas with sparse data rely on estimates produced using models that extrapolate trends from other regions or time periods. This talk describes the statistical challenges inherent in relying on such extrapolation for policy decision-making. Using data from verbal autopsies (surveys conducted to ascertain a likely cause of death in settings without routine death registration), this talk will explore the trade-offs between statistical/machine learning algorithms and choice of training data, before ending with preliminary work on visualization tools to communicate uncertainty from extrapolation to policymakers.

No extended abstract or paper available

 Presented in Session 203. Demographic Estimation for Monitoring and Decision-making in Sparse-Data Settings