Understanding GPU programming for statistical computation: Studies in massively parallel massive mixtures

This article describes advances in statistical computation for large-scale data analysis in structured Bayesian mixture models via graphics processing unit (GPU) programming. The developments are partly motivated by computational challenges arising in fitting models of increasing heterogeneity to increasingly large datasets. An example context concerns common biological studies using high-throughput technologies generating many, very large datasets and requiring increasingly high-dimensional mixture models with large numbers of mixture components.We outline important strategies and processes for GPU computation in Bayesian simulation and optimization approaches, give examples of the benefits of GPU implementations in terms of processing speed and scale-up in ability to analyze large datasets, and provide a detailed, tutorial-style exposition that will benefit readers interested in developing GPU-based approaches in other statistical models. Novel, GPU-oriented approaches to modifying existing algorithms software design can lead to vast speed-up and, critically, enable statistical analyses that presently will not be performed due to compute time limitations in traditional computational environments. Supplementalmaterials are provided with all source code, example data, and details that will enable readers to implement and explore the GPU approach in this mixture modeling context. © 2010 American Statistical Association, Institute of Mathematical Statistics, and Interface Foundation of North America.

Prevalence and predictors of giving birth in health facilities in Bugesera District, Rwanda.

BACKGROUND: The proportion of births attended by skilled health personnel is one of two indicators used to measure progress towards Millennium Development Goal 5, which aims for a 75% reduction in global maternal mortality ratios by 2015. Rwanda has one of the highest maternal mortality ratios in the world, estimated between 249-584 maternal deaths per 100,000 live births. The objectives of this study were to quantify secular trends in health facility delivery and to identify factors that affect the uptake of intrapartum healthcare services among women living in rural villages in Bugesera District, Eastern Province, Rwanda. METHODS: Using census data and probability proportional to size cluster sampling methodology, 30 villages were selected for community-based, cross-sectional surveys of women aged 18-50 who had given birth in the previous three years. Complete obstetric histories and detailed demographic data were elicited from respondents using iPad technology. Geospatial coordinates were used to calculate the path distances between each village and its designated health center and district hospital. Bivariate and multivariate logistic regressions were used to identify factors associated with delivery in health facilities. RESULTS: Analysis of 3106 lifetime deliveries from 859 respondents shows a sharp increase in the percentage of health facility deliveries in recent years. Delivering a penultimate baby at a health facility (OR = 4.681 [3.204 - 6.839]), possessing health insurance (OR = 3.812 [1.795 - 8.097]), managing household finances (OR = 1.897 [1.046 - 3.439]), attending more antenatal care visits (OR = 1.567 [1.163 - 2.112]), delivering more recently (OR = 1.438 [1.120 - 1.847] annually), and living closer to a health center (OR = 0.909 [0.846 - 0.976] per km) were independently associated with facility delivery. CONCLUSIONS: The strongest correlates of facility-based delivery in Bugesera District include previous delivery at a health facility, possession of health insurance, greater financial autonomy, more recent interactions with the health system, and proximity to a health center. Recent structural interventions in Rwanda, including the rapid scale-up of community-financed health insurance, likely contributed to the dramatic improvement in the health facility delivery rate observed in our study.