Math@Noon Fridays

Math@Noon is a weekly event featuring presentations on math research topics and workshops on computational technologies, opportunities for students and recreational mathematics. Topics vary each week. Open to anyone who enjoys math! We meet on Fridays at noon in KEP 3160.

Seminar Talk Abstract:

A Computationally Efficient Poisson GLM Spatial Scan Method

Kayley Smiley and Joshua P. French

The spatial scan method has been widely used to detect disease clusters by identifying zones where the observed case count is significantly higher than expected. While the original method only uses basic information, such as case and population counts, incorporating additional explanatory information related to demographics, the environment, and health can dramatically impact the expected case count. An extension using generalized linear models (GLMs) allows such information to be included, but it has not gained much traction in practice due to high computational cost. We present a computationally efficient version of the Poisson GLM spatial scan method that drastically reduces runtime compared to the original, naïve implementation. We will introduce the original scan method, explain its extension to include explanatory variables, and describe how the method can be implemented efficiently. We will then illustrate the utility of the new method through a simulation study and an application to real public health data.