Colloquium

Department of Statistics

Statistics Colloquium
April 22 (Fri) 11 a.m.
Alavi Commons Room, 6625 Everett Tower

Latent cluster modeling of spatio-temporal variation in small area health data: an application to the smoking ban in Glasgow, Scotland

Andrew B. Lawson
Medical University of South Carolina

In this talk I present a novel approach to spatio-temporal cluster modeling for small area health data. The basic model is developed for counts within small areas and fixed time periods and consists of a relative risk model where fixed mixture risk components are allowed to have spatial and temporal structure. The mixture model is sampled via conventional posterior sampling algorithms with variants of Metropolised Carlin-Chib for mixture components. Three different spatial prior models are considered and monotone risk levels are assumed. Temporal clustering components can be assigned to aggregated spatial regions (clusters) and inference on changepoints in time can be made. An example of smoking cessation policies in Scotland and respiratory health will be explored.
~
All statistics students are expected to attend.

Past colloquiums