COMPSTAT 2016: Start Registration
View Submission - CRoNoS FDA 2016
A0183
Title: Functional regression for investigating the feeding behavior of animals Authors:  Jan Gertheiss - Clausthal University of Technology (Germany) [presenting]
Sonja Greven - LMU Munich (Germany)
Engel Hessel - University of Goettingen (Germany)
Verena Maier - LMU Munich (Germany)
Fabian Scheipl - Ludwig-Maximilians-Universitaet Muenchen (Germany)
Abstract: A group of pigs is observed 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 sequences of binary observations for each pig and day. Goals of the data analysis are to find pig-specific feeding profiles showing the typical feeding pattern of each pig, and to make short-term predictions of pig-specific feeding probabilities. Different approaches for modeling the data are discussed: (1) a marginal functional logistic regression approach modeling the binary measurements by assuming latent, smooth and cyclic pig-specific profiles. To account for correlation of measurements, robust standard errors and corresponding pointwise confidence intervals can be used. As an alternative, (2) a conditional model including pig-specific functional random effects or lagged responses to account for within-pig correlation is considered. By contrast to the marginal model, the latter model also allows for short-term predictions of feeding behavior.