|
Activity Number:
|
271
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistical Graphics
|
| Abstract - #305172 |
|
Title:
|
Voronoi Tessellation for Contour Plots
|
|
Author(s):
|
Hyunsook Lee*+
|
|
Companies:
|
Harvard-Smithsonian Center for Astrophysics
|
|
Address:
|
60 Garden St. MS70, Cambridge, MA, 02138,
|
|
Keywords:
|
Voronoi tessellation ; contours ; nonparametric density estimation ; algorithm ; mode identification ; high energy astrophysics
|
|
Abstract:
|
Contour plots are the most frequently adopted display method when data points are quite populated and hardly distinguishable in scatter plots. In order to get contours of certain percentiles, often imposed are parametric models and distributions including kernel functions. Their shortcomings are posing ellipsoid shape contours, and expensive but slow computation with massive data. Overcoming such obstacles has been a consistent demand in astronomy owing to colossal data size. Here, we propose Voronoi tessellation for contouring and enhancing information of scatter plots. The tessellation does not impose any parametric density functions. As opposed to kernel density estimation, it is computationally efficient while delineating the shape of data distributions. The algorithm is illustrated with simulated data and applied to scatter plots from X-ray spectral data analysis.
|
