Документ взят из кэша поисковой машины. Адрес
оригинального документа
: http://hea-www.harvard.edu/AstroStat/Stat310_fMMIV/
Дата изменения: Unknown Дата индексирования: Mon Oct 1 22:38:18 2012 Кодировка: |
This is a broad overview of astronomy as an introduction to Astrostatistics. We describe research methods and techniques, as well as questions posed by some recent observations. We concentrate on problems related to high energy astrophysics and consider the data obtained by modern X-ray space telescopes such as Chandra X-ray Observatory and XMM-Newton. In this first lecture we define basic astronomical terms and present different type of data used by astrophysicists. We also describe statistical methods applied in the standard data analysis process. We show examples of the data analysis and physical interpretation of the results.
The atomic and nuclear age of the past century allowed astronomers (and others) to peer into the "invisible world" of high- and low- energy radiation, from radio waves to gamma-rays, to see and understand the underlying quantum- and relativistic processes. How did the inference techniques change and grow with the new sciences? In this presentation, we go over the growth (and stagnation, sometimes!) of statistics in modern astrophysics; from the eye-and-hand techniques to more sophisticated imaging, energy-spectra, and timing methods. We will have some of the early instrumentation, and stories from that time, on hand -- as well as models of new twenty-first century instruments.
The foundations of astrophysics are rooted in photon spectra, and spectral energy distributions are generally our sole source of information on the composition and environment of extrasolar objects. We will introduce concepts such as atomic lines, line and continuum emission, ionization balance, and emission measure distributions, with emphasis on how they are used to understand the coronae of solar like stars.
I will describe an extension of the standard wavelet-based estimation framework to the class of generalized linear models, based on the use of recursive partitioning and piecewise polynomials. Estimates produced in this setting yield information on both scale and extent of local structure in an underlying time series. They are accompanied by both efficient algorithms and near-optimal theoretical properties. I will illustrate the use of these models in the context of estimating flux underlying gamma-ray burst signals. Numerous extensions are possible.
I will continue to talk about the on-going project for estimating the distribution of the temperature of stellar corona also known as the differential emission measure (DEM).
I will focus on hierarchical missing data structuring and handling the emissivity matrices in terms of efficient data augmentation.
Correcting the wavelength errors at ATOMDB will be discussed. I will provide the model diagnostics related to the error correction by using posterior predictive distribution.