|
Activity Number:
|
284
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 14, 2002 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Stat. Sciences*
|
| Abstract - #301733 |
|
Title:
|
Bayesian Analysis of Compositional Time Series by Using Multivariate Skewed Normal Distribution
|
|
Author(s):
|
Rongwei Fu*+ and Dipak Dey and Nalini Ranvishanker
|
|
Affiliation(s):
|
University of Connecticut and University of Connecticut and University of Connecticut
|
|
Address:
|
U-4120, 215 Glenbrook Road, Storrs, Connecticut, 06269, U.S.A
|
|
Keywords:
|
compositional time series ; multivariate skewed normal distribution ; additive logistic ratio transformation ; Bayesian analysis
|
|
Abstract:
|
There has been a general tendency in the stastistical literature towards more flexible methods to represent features of the data as adequately as possible and reduce unrealistic assumptions. While the asssumption of normal distribution has played an overwhelming role in the multivariate analysis, multivariate skewed normal distribution could be used as one alternative and represents a mathematically tractable extension of the miltivariate normal density with the addition of a parameter to regulate skewness. Compositional time series are unit-sum constrained multivariate time series with important application in desciplines such as geology, economics, and ecology, but accurate inference is often difficult due to lack of suitable classes of parametric distribution. In this study, multivariate skewed normal distribution are used to model additive logistic ratio (ALR) transformed compostional data within a hierarchical Bayesian framework. Inference could be achieved by using Markov chain Monte Carlo techniques.
|
