Marginal Functional Regression Models for Analyzing the Feeding Behavior of Pigs

• Published in: Journal of agricultural, biological, and environmental statistics Vol. 20; no. 3; pp. 353 - 370
• Main Authors: Gertheiss, Jan, Maier, Verena, Hessel, Engel F, Staicu, Ana-Maria
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.09.2015; Springer Science+Business Media, LLC; Springer
• Subjects: Agriculture; Analysis; Biostatistics; confidence interval; feeding behavior; Health Sciences; humidity; Mathematics and Statistics; Medicine; Monitoring/Environmental Analysis; radio frequency identification; Radio frequency identification (RFID); radio waves; regression analysis; Statistics; Statistics for Life Sciences; Swine; swine feeding; temperature; Animal husbandry; Pig fattening; Penalized splines; Binary functional data; Generalized additive models
• ISSN: 1085-7117; 1537-2693
• DOI: 10.1007/s13253-015-0212-7

We observe a group of pigs over a period of about 100 days. Using high frequency radio frequency identification, it is recorded when each pig is feeding, leading to very dense binary functional data for each pig and day. One aim of the data analysis is to find pig-specific feeding profiles showing us the typical feeding pattern of each pig. For modeling the data, we use a marginal functional logistic regression approach, allowing us to model the densely observed binary measurements by assuming an underlying smooth subject-specific profile. The method also allows to incorporate additional covariates such as temperature and humidity that may influence the pigs’ behavior. To account for correlation of measurements, we use robust standard errors and corresponding pointwise confidence intervals. Before analyzing the feeding behavior of pigs, the method employed is evaluated in simulation studies. As our approach is rather general, it may also be applied to other types of generalized functional data with similar characteristics as the pig data.