Modeling Binary Functional Data with Application to Animal Husbandry

• 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): Jan Gertheiss*+ and Verena Maier and Engel F. Hessel and Ana-Maria Staicu
• Companies: Georg August University and Ludwig Maximilians University and Georg August University and North Carolina State University
• Keywords: categorical functional data ; generalized additive models ; marginal models ; pig fattening

Abstract

We observe a group of pigs over a period of about 100 days. On a very dense grid of time points, it is recorded when each pig is feeding, leading to binary functional data for each pig and day. In addition, there are measurements such as temperature and humidity available that may influence the pigs' behavior. One aim of the data analysis is to find pig-specific feeding profiles telling us when a certain pig is typically feeding. For analyzing the data, we propose a functional logistic regression approach allowing us to model the binary but functional measurements by assuming an underlying smooth pig-specific profile. The method also allows to incorporate additional (potentially non-functional) covariates. The profiles we obtain can then be used for clustering/comparing pigs, or we may try to relate the profiles to the weight of the pig. Though the approach presented has originally been designed for analyzing the pig data above, it is rather general and hence applicable to other types of binary functional data, too.