Antipodally Symmetric Distributions for Orientation Statistics.
Abstract
The conventional antipodally symmetric Bingham matrix distribution on the Stiefel manifold is generalized. Large sample maximum likelihood estimation and uniformity tests are discussed, and a parametric model for axial orientations (X-shapes) is suggested. A generalization of the Khatri-Mardia matrix distribution is developed to provide a model suitable for hybrids (T-shapes). Beran's results on exponential models for directional data are extended to orientation statistics to provide regression estimators and goodness-of-fit tests as alternatives to maximum likelihood estimation and likelihood ratio tests. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1981
- Accession Number
- ADA097460
Entities
People
- Michael J. Prentice
Organizations
- Princeton University