Ordinal and Nominal Functional Data Analysis

• Activity Number: 162
• Type: Topic Contributed
• Date/Time: Monday, August 4, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Bruce Swihart*+
• Companies: Johns Hopkins Bloomberg School of Public Health
• Keywords: functional linear model ; linear mixed model ; zero-variance component testing

Abstract

Functional principal components decompositions are a typical tool used in functional data analysis, where the functions decomposed are continuous. Sleep hypnograms are discrete-state discrete-time step-functions describing the sleep process and have recently been modeled with Poisson regression and multi-state survival analysis methods, which unprecedentedly honor the transition and temporal information but fail in representing the entire trajectory. To account for more information, an applied technique for such nominal functional data is introduced and used in the context of scalar-on-function regression and compared to scalar-on-scalar regressions of the Poisson regression features.