|
Activity Number:
|
436
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 12, 2004 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #301056 |
|
Title:
|
Bayesian Estimation of Generalized Lambda Distributions Using MCMC
|
|
Author(s):
|
Michele A. Haynes*+ and Robert King and Kerrie Mengersen
|
|
Companies:
|
University of Queensland and University of Newcastle and University of Newcastle
|
|
Address:
|
UQ Social Research Centre, Brisbane, 4072, Australia
|
|
Keywords:
|
generalized lambda distributions ; Bayesian estimation ; MCMC ; posterior distributions
|
|
Abstract:
|
The Generalized Lambda Distributions (GLD), a family of flexible distributions defined through a quantile function, have previously been estimated using moment-based methods. Recently, a transformation-based procedure for estimation known as the starship method has been shown to perform favorably. This procedure does not rely on computation of the likelihood but focuses on how well the reversely transformed data fits a base distribution (in the case of the GLD, the uniform). We investigate a Bayesian procedure for the estimation of the FMKL-parameterization of the GLD which is well defined for all values of the two shape parameters. MCMC provides a simulation-based alternative for deriving posterior estimates of the parameters through a Bayesian framework. This approach is also advantageous as it allows us to place sensible prior distributions on the parameters to govern the shape of the posterior distributions during the simulation process. We simulate sample data from symmetric and asymmetric distributions and estimate parameters for the GLD using both MCMC and starship estimation procedures for comparison.
|
