|
Activity Number:
|
120
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #304275 |
|
Title:
|
Parameter Estimation in Two-Sample Location-Scale Families
|
|
Author(s):
|
Cornelis Potgieter*+ and Fred Lombard
|
|
Companies:
|
University of Johannesburg and University of Johannesburg
|
|
Address:
|
, , International, , South Africa
|
|
Keywords:
|
location-scale families ; parameter estimation ; quantiles ; asymptotic likelihood
|
|
Abstract:
|
The difference between two populations belonging to the same (but unspecified) location-scale family can be characterized by two parameters. When samples are observed from each of the populations, there are several methods for estimating these parameters, including a method of moments (MOM), a method of quantiles (MOQ) and a generalized least squares method (GLS). A new asymptotic likelihood method (AL) is also proposed here. The GLS and AL methods are based on a fixed number of quantiles. A plug-in approach for choosing these quantiles is discussed. Asymptotic efficiencies of the four methods of estimation are compared. The results from a Monte Carlo study provide insight into the small-sample behavior of the four estimators.
|
