2018-01-26

Soutik Ghosal, PhD Candidate, Department of Bioinformatics and Biostatistics, University of Louisville

"Spatiotemporal mixed effect modeling and its applications"

Spatiotemporal analysis is a very popular and effective tool to identify variation over both space and time. It has widespread applications on various fields such as meteorology, ecology, public health and medicine, and biology. Spatiotemporal analysis, a broad evolving research area, constitutes of numerous statistical techniques. In this talk, I will discuss about two kinds of spatiotemporal analysis tools to address two different data problems. In the first case, we are interested to study health professional shortage areas in the United States where the number of primary care physicians is considered as the outcome variable, and is collected at county level and over different years. The outcome variable is a zero-inflated count data and county level characteristics (e.g. population size, education level, and poverty level) may impact the number of doctors in that area. Since the data is repeatedly collected over time, counties are nested within the state, and adjacent counties are geographically correlated, the dependence structure of the data is very complex. We develop a Bayesian mixed effect hurdle model as a spatiotemporal analysis tool to model this data. In the second study, we model the overuse of antibiotic prescriptions for upper respiratory tract infection (URTI) diagnosis using relevant demographic and socio-economic covariates and model the spatial heterogeneity in zip code level within Kentucky and the seasonal variation within the years 2014-2016. To do these, we incorporate thin plate spline to address the spatial heterogeneity of antibiotic prescriptions and trend and seasonal components to assess the temporal variation in a generalized additive logistic regression model.

Stay connected TwitterFacebookLinkedInYouTubeInstagram