|
Activity Number:
|
289
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #307429 |
|
Title:
|
A Nonparametric Approach to Descriptive Measures of Multivariate Massive Data Based on Convex Hull Peeling Depth
|
|
Author(s):
|
Hyunsook Lee*+
|
|
Companies:
|
The Pennsylvania State University
|
|
Address:
|
265 Blue Course Drive, Apt. 23B, State College, PA, 16803,
|
|
Keywords:
|
convex hull peeling ; descriptive statistics ; nonparametric multivariate analysis ; massive data
|
|
Abstract:
|
Visualizing multidimensional data is often biased due to projection on less dimensions. Quantizing shapes of multivariate data is generally limited to the normal distribution assumption. In this presentation, we introduce descriptive statistics to measure the skewness and kurtosis of multidimensional data based on convex hull peeling depth. These convex hull peeling algorithms do not require calculating moments but provide a mapping to one dimensional scale curve, characterizing shapes of multivariate distributions. As diagnostic tools, convex hull peeling algorithms are exemplified with Monte Carlo simulations and data sets from Sloan Digital Sky Survey.
|
