Presented By: Department of Statistics Dissertation Defenses
Bayesian Perspectives on LongROAD Study: Analyzing Driving Decline and Latent Traits
Vincenzo Loffredo
Abstract: The LongROAD study presents several challenges in analyzing and interpreting the data collected. In this thesis, we discuss the use of Bayesian Nonparametric methods using Gaussian Processes to analyze the evolution of trends in a variable, with an application to how aging affects individuals' driving behaviors. We introduce the use of the velocity and the velocity field framework to improve prediction outputs for the evolution of the time series. Moreover, we provide a Bayesian parametric framework based on Beta distributed latent traits to model the output of Likert Scale surveys and capture the distribution of the latent traits of interest at the population level. Finally, we describe an extension of the velocity and velocity field framework to integrate covariates collected at different frequencies and develop an interpretable model to assess which factors affect more the evolution of a variable of interest.
Explore Similar Events
-
Loading Similar Events...