|
Activity Number:
|
376
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #310376 |
|
Title:
|
Some New Models and Procedures for Directional Data
|
|
Author(s):
|
Brett Presnell*+ and Pavlina Rumcheva
|
|
Companies:
|
University of Florida and C-BASS, LLC
|
|
Address:
|
Dept of Statistics, Gainesville, FL, 32611-8545,
|
|
Keywords:
|
Directional data ; Langevin distribution ; von Mises distribution ; Fisher distribution ; Projected normal distribution ; Random effect
|
|
Abstract:
|
Some new inferential procedures are presented for directional data in the general d-dimensional case, with two and three dimensional directions being of primary interest. We first consider new tests for equality of mean directions and concentrations for multisample data from the Langevin-von Mises-Fisher and the projected normal distributions. The tests for equality of mean directions in particular perform well over broad ranch of concentrations, reducing or eliminating the need to tailor the test to the underlying unknown concentration. We also discuss and fit random and mixed effects models for directional data. These models generalize the probit-normal generalized linear mixed model from the binary case of a one dimensional direction to higher dimensions.
|
