|
Activity Number:
|
85
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2003 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Stat. Sciences
|
| Abstract - #301977 |
|
Title:
|
Bayesian Methods for Circular Time Series Using Wrapped Distributions
|
|
Author(s):
|
Palanikumar Ravindran*+ and Sujit Kumar Ghosh
|
|
Companies:
|
Roche and North Carolina State University
|
|
Address:
|
1000 Kiely Blvd. Apt. 73, Santa Clara, CA, 95051,
|
|
Keywords:
|
time series ; circular data ; wind direction ; Bayesian ; wrapped normal ; data augmentation
|
|
Abstract:
|
Circular time series data arise in a number of different areas such as meteorological and oceanographical sciences. A popular example is the hourly or daily measurements of wind direction. We cannot use standard statistical techniques to model circular data, due to the circular geometry of the sample space. We use the wrapping approach, which assumes that the generating distribution for circular data is obtained by wrapping a distribution on the real line onto a circle. This approach creates a huge class of circular data models. However the likelihood-based inference for such distributions can be very complicated and computationally intensive. This talk discusses how Markov chain Monte Carlo using data augmentation techniques overcome these difficulties. It is assumed that the wrapping coefficients are missing and are used along with the observed directions to fit the complete model. We can easily compute the usual summary statistics and other statistics such as circular autocorrelation functions and their standard errors. We apply our method to a wind direction dataset and compare our results to parameter estimates available in the literature.
|
