19
T.W. Alexander Drive P.O.Box 14006 Research Triangle Park, NC 27709-4006 Tel: 919.685.9350 Fax: 919.685.9360 [email protected] |
Astrostatistics Opening Workshop
Poster Session (1/23/06, 6:30-8:30 p.m.)
Abstracts
Estimating Deformations of Isotropic Gaussian Random Fields
on the Plane.
Ethan Anderes (U.C.-Berkeley, Dept. of Statistics)
We present a new approach to the estimation of the deformation of an isotropic
Gaussian random field on $\Bbb R^2$ based on dense observations of a single
realization of the deformed random field. Under this framework we investigate
the identification and estimation of deformations. We also present a complete
methodological package---from model assumptions to algorithmic recovery of the
deformation---for the class of non-stationary processes obtained by deforming
isotropic Gaussian random fields.
Inferring Galaxy Morphology Through Texture Analysis
Kinman Au (Carnegie Mellon University, Dept. of Statistics)
We give an approach to estimate galaxy morphology from digital images. In particular,
our algorithm extracts orientation information of the texture at difference
scales, and merges the multiscale information into an unified representation.
By fitting a morphological model based on the textural information, we derive
an quantitative and physically meaningful description of galaxy morphology.
Such description will help scientists to study how galaxy morphology evolve
over time, and the effect of environment toward the evolution. The answers will
provide important clues about the origin of the Universe.
Multivariate Analysis of Gamma-Ray Bursts from BATSE 4B
Ruth-Stella Barrera-Rojas and Antonio Uribe (Universidad
Nacional de Colombia)
We present here a multivariate analysis of gamma-ray burst (GRB) properties for discriminating between distinct classes of GRBs. We are applying the methodology proposed by (Mukherjee, Feigelson et al, 1998), so we use a multivariate clustering procedure on a sample of 1637 bursts from the Forth BATSE Catalog, this is a parametric maximum likelihood model-based clustering procedure assuming multinormal populations calculated with the EM Algorithm and validated with the Bayesian Information Criterion, the software for the calculations is EMMIX.
Finding Smaller Exoplanets
Floyd Bullard (Duke, ISDS and SAMSI)
Arbitrarily small exoplanets can be detected using radial velocity measurements even when the velocity measurements' margins of error are substantially greater than the velocity function's amplitude. The profile log likelihood function will always show a clear peak at the true period with enough observations.
The Wide-field Imaging Nearby Galaxy Survey
Antonio Cava (Milano University & INAF-OAPD)
A Comparison of Least-Squares and Bayesian Techniques in Fitting
the Orbits of Extrasolar Planets
Peter Driscoll (San Francisco State University)
High Energy Photon Emission in Young Supernova Remnants
Don Ellison, Diana Yanchukova (both -- NCSU, Dept. of
Physics)
Young SNRs are believed to produce cosmic ray ions and electrons, but direct
evidence for ion acceleration in SNRs remains illusive. An important key to
the solution of this problem concerns the relative efficiency for producing
inverse Compton radiation vs. gamma-ray production via pion decay. We outline
elements of this problem and compare results to recent H.E.S.S. TeV observations
of young SNRs.
Developing a Bayesian Toolbox for Detection and Orbit Determination
of Extrasolar Planets
Eric Ford (Univ. of California - Berkeley)
Radial velocity surveys have now detected over 150 extrasolar planets around nearby main sequence stars. Many of these planets are clearly detected and have well characterized orbits, thanks to a large ratio of the velocity amplitude to measurement precision and observations spanning many orbital periods. However, a growing number of planets have orbital periods comparable to the duration of observations and/or induce radial velocity variations not much larger than the measurement precision (Fig. 1). For such planets, there are often large uncertainties in the orbital parameters. In the most extreme cases, even establishing the reality of a periodic signal is difficult. These difficulties become even more severe for multiple planet systems which require simultaneously fitting numerous model parameters. So far, most analyzes of extrasolar planets have relied on frequentist methods such as maximum likelihood. I review recent progress in developing the necessary computational tools for implementing such analyzes. I demonstrate these techniques with a Bayesian analysis of a recent triple planet system orbiting HD 37124.
Non-Parametric Analysis of Supernova Data and the Dark Energy
Equation of State
and
Examining the Effect of the Map-Making Algorithm on Observed Power
Asymmetry in WMAP Data
Peter Freeman (Carnegie Mellon University)
A Bayesian Multi-Planet Kepler Periodogram for Exoplanet Detection
Philip Gregory (University of British Columbia)
Density Estimation and Clustering\\ in Astronomical Sky Surveys
Woncheol Jang (Duke)
The Universe is homogeneous and isotropic on large scales, but on small scales, one can find significant deviations from homogeneity and isotropy. Clusters of galaxies play an important role in finding where the local structure fades away into a homogeneous and isotropic distribution. From statistical point of view, finding clusters of galaxies is equivalent to finding density contour clusters (Hartigan, 1975), connected components of a level set Sc ={f > c}. We present a nonparametric method to find density contour clusters. To extract connected components of the estimated level est, we propose to use a union of balls to approximate the estimated level set which is a modified version of Cuevas et al (2000). The method is applied to studies of the Edinburgh-Durham Southern Galaxy Catalogue (EDSGC) and Mock 2df catalogue. Results are compared with the existing galaxy catalogs.
A Bayesian Approach to Analyzing Star Cluster Parameters
William Jefferys (Universities of Texas and Vermont)
Convex Hull Peeling: Nonparametric Multivariate Data Analysis
Tools
Hyunsook Lee (Penn State)
An ad hoc device on multidimensional massive data is in demand. However, multivariate
data analysis tools not imposing multivariate normal distribution exist rarely.
We introduce convex hull peeling algorithms as a such device for the analysis
of multidimensional massive data. Only the convexity of data sets is assumed.
These convex hull peeling algorithms are designed to estimate quantiles, detect
outliers, and measure distribution shapes of multidimensional data. Additionally,
the algorithms are exemplified with Monte Carlo simulations and SDSS DR4 Quasars.
Estimating a Decreasing Density for the Dark Matter in Nearby
Dwarf Galaxies
Jayanta Pal (University of Michigan)
Efficient X-ray Spectral Fitting with Narrow Emission Lines
Taeyoung Park (Harvard University, Dept. of Statistics)
From a statistical point of view, spectral analysis is the modeling of the distribution
of photon energies, a distribution that can be formulated as a finite mixture
of two photon groups, a continuum term and a set of emission lines. While the
continuum describes a general shape of a spectrum, each emission line represents
a positive aberration from the continuum in a narrow band of energies. Here,
we focus on a single emission line that can be modeled with a Gaussian distribution
or a delta function. Spectral data are contaminated by several non-trivial physical
processes including non-homogeneous stochastic censoring, blurring of photon
energies, and background contamination. To account for these processes, we consider
a hierarchical structure of missing data under a Bayesian perspective. To fit
the resulting highly structured multilevel spectral models, we devise efficient
Gibbs sampling strategies. As an illustration, we apply our strategies to the
X-ray spectrum of the high redshift quasar, PG 1634+706.
Characterization of Galaxy Evolution as function of Local
Environment
Alex Rojas (Carnegie Mellon University)
Highly Structured Models and Statistical Computation in High-Energy
Astrophysics
Adam Roy (University of California - Irvine)
Genetic Algorithms for Gravitational Lenses
David Valls-Gabaud (Canada-France-Hawaii Telescope)
Entire site © 2001-2005, Statistical and Applied Mathematical Sciences Institute. All Rights Reserved.
This page updated on January 20, 2006 11:05 AM