B0404
Title: Continuous glucose monitoring using distributional regression models
Authors: Jenifer Espasandin-Dominguez - University of Santiago de Compostela (Spain) [presenting]
Carmen Cadarso Suarez - Universidad de Santiago de Compostela (Spain)
Thomas Kneib - University of Goettingen (Germany)
Francisco Gude - Complexo Hospitalario Universitario de Santiago de Compostela (Spain)
Abstract: The technological progress has led to the development of new measurement procedures in the form of functional data. We propose to incorporate this functional information within the framework of the distributional regression models. This type of models are a generic framework for performing regression analyses where every parameter of a potentially complex response distribution is related to an additive predictor. In the Bayesian inference framework, structured additive distributional regression models extend the use of generalized additive models to situations in which the response distributions are nonstandard, and where not only the mean but multiple parameters are related to additive predictors. Further, they allow additional flexibility by specifying structured additive predictors for each parameter of interest, and thus adjust for several types of covariate effects. The methodologies developed will be applied to real biomedical data, in a study of glycated haemoglobin, a test useful in the control of individuals with diabetes. The predictor will include the results of continuous monitoring, which collects glucose measurements every 5 minutes over a week. The glucose levels will be included as a functional covariate in a scalar on functional regression model. The inclusion of glucose profiles as a predictor of this type of models will mark a novel advance in the study of diabetes.
