|
Activity Number:
|
509
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 10, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Quality and Productivity
|
| Abstract - #306207 |
|
Title:
|
Adaptively Trimmed L-Moments with Applications to Heavy-Tailed Distributions
|
|
Author(s):
|
Jonathan Hosking*+
|
|
Companies:
|
IBM Research
|
|
Address:
|
P.O. Box 218, Yorktown Heights, NY, 10598,
|
|
Keywords:
|
estimation ; heavy-tailed distributions ; l-moments ; order statistics
|
|
Abstract:
|
L-moments are statistics that enable summarization of data samples and parameter estimation of probability distributions using linear combinations of order statistics. Elamir and Seheult (2004) introduced trimmed L-moments, which exclude one or more extreme order statistics and can be used for inference about heavy-tailed distributions for which the ordinary L-moments may not exist. However, the appropriate degree of trimming is usually not clear a priori. Here I define adaptively trimmed L-moments, in which the degree of trimming is suggested by the data values themselves. In particular for the generalized Pareto distribution, a simple relation between expectations of fractional order statistics can be used to derive estimators that remain valid no matter how heavy the tail of the distribution. I define these estimators and explore some of their theoretical and practical properties.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2006 program |
