Sudipto Banerjee – Space, Time and Gradients: Why We Need Them in Statistical Modeling for Public Health Data

Click to download Dr. Banerjee's slides (168)

CHIPTS Methods Seminar – UCLA-Semel Institute Center for Community Health

Sudipto Banerjee, PhD
Professor and Chair
Dept. of Biostatistics
UCLA Fielding School of Public Health

Tuesday, February 3, 2-3pm
Center for Community Health, UCLA Wilshire Center
10920 Wilshire Blvd., Suite 350, Conference Room

Advances in Geographical Information Systems (GIS) and related software have led to a burgeoning of spatial-temporal databases. Statisticians and spatial analysts today routinely encounter situations where they seek to model relationships among variables across space and time. In recent times interest has turned to inferring about rates of change of health outcomes over space and time. Why are such questions relevant and how should we estimate them? One example considers analyzing monthly hospitalization rates aggregated over the counties in California where hospital management seeks to carry out inference on gradients of the temporal process, while at the same time accounting for spatial similarities across neighboring regions. Another example (an extension) is to analyze spatial-temporal gradients for environmental pollutants to understand the nature of dispersal of pollutants. Here, we are interested in directional rates of change over space at any given time, temporal gradients at any given location and even “mixed” gradients, e.g., how the temporal rate of change varies over space. We will work within a fully Bayesian inferential paradigm without unnecessary, and potentially inflexible, parametric modeling assumptions and obtain the full posterior predictive distribution for these gradients using process-based models.

(Co-authors: Harrison Quick and Bradley P. Carlin)