Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 208 - Survey Estimation
Type: Contributed
Date/Time: Tuesday, August 4, 2020 : 10:00 AM to 2:00 PM
Sponsor: Section on Bayesian Statistical Science
Abstract #313690
Title: Imposing Sparseness in a Bayesian Hierarchical Regression Model with Temporal Smoothing via the Horseshoe Prior with an Application to Estimate Stillbirths for All Countries
Author(s): zhengfan Wang* and Leontine Alkema and Miranda Fix and Jon Wakefield and Hannah Blencowe and Lucia Hug and Danzhen You and Anupam Mishra
Companies: umass-Amherst and University of Massachussetts and University of Washington and Departments of Biostatistics and Statistics, University of WAshington and LSHTM and UNICEF and UNICEF and UNICEF
Keywords: hierarchical Bayesian model; smoothing process; sparsity ; Horseshoe prior; B-splines
Abstract:

Estimation of stillbirth rates (SBRs) globally is complicated because of the absence of reliable data from countries where most stillbirths occur. We compiled data and developed a hierarchical Bayesian regression model for estimating stillbirth rates for all countries from 2000 to 2020. The model combines covariates with a temporal smoothing process (B-spline) such that estimates are data-driven in country-periods with high-quality data and determined by covariates for country-periods with limited or no data. Horseshoe priors are used to shrink the irrelevant covariates to zero. The model accounts for bias and additional uncertainty associated with observations with alternative stillbirth definitions and observations that are subject to non-sampling errors. In-sample goodness of fit and out-of-sample validation results suggest that the model is well calibrated and produces unbiased estimates. The model is used by UN agencies to monitor the stillbirth rate for all countries.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program