B0870
Title: Multimodality tests for circular data
Authors: Jose Ameijeiras-Alonso - Universidade de Santiago de Compostela (Spain) [presenting]
Rosa Crujeiras - University of Santiago de Compostela (Spain)
Alberto Rodriguez-Casal - University of Santiago de Compostela (Spain)
Abstract: In applied statistics, the identification of the number of preferred directions in animal movements or the assessment of peak times in data observed across time, can be reduced to the same statistical problem: determining the number of modes (local maxima of the probability density function) in the underlying distribution of a sample of angles. Testing for multimodality provides a formal way to determine if the (circular) distribution of the sample has a specific number of modes. Within this context, there are various proposals, both for scalar and for circular data. However, the poor calibration of the methods provides unsatisfactory results in practice. A method for assessing multimodality in circular data will be presented, jointly with a calibration algorithm whose performance will be supported by simulation experiments and a real data application.
