Flexible Multi-dimensional Modeling of Complex Data in Astronomy
229th AAS Meeting January 4, 2016 9:30 - 11:30 am Gaylord Texan Resort & Convention Center Grapevine 4
Updated: The presentation slides have been uploaded to the site. Please feel free to ask us questions at the Chandra booth throughout the week.
Recent improvements in instrumentation and the data collection process across the entire electromagnetic spectrum have resulted in an increasing amount of high quality multi-wavelength observations. The analysis of these modern data sets present several statistical challenges that require new methods and techniques to support the scientific inference. Our session will focus on the discussion of applied methodology that can be used to tackle some of these challenges. We will present tutorials based on the Sherpa-Python and IRIS tools developed by the Chandra X-ray Observatory.
Sherpa is a Python based general modeling and fitting application that provides an environment for modeling multi-dimensional data with a set of optimization methods, including MCMC simulations for sampling posterior distributions. Sherpa provides flexible mechanisms for modeling Poisson (sparse) and Gaussian (rich) data with appropriate likelihoods, including both pre-defined models and an interface to incorporate user defined models (Python functions or external code). Sherpa can be used for modeling 1D, 2D, or 3D data, i.e., spectra, time-series, or images, and can be extended to spectral-timing and spatial-timing domains. An upcoming 'Sherpa to Astropy' Python package will allow users to use Sherpa's optimizers and error estimators seamlessly within the Astropy's modeling framework. Iris has been built on top of Sherpa for fitting SEDs to multi-wavelength data. Iris also provides a front-end to Virtual Observatory archival catalogs that can supply the appropriate data for the modeling session.
We will use IPython Notebooks to guide the participants through Sherpa-Python sessions and present a tutorial demonstration showing Iris connectivity to the archives and examples of SED modeling.
There will be free time for participants to explore any of the several science threads provided at the end of the demonstrations, or to work on their own analysis. The session leaders will be available for help and answering any questions participants have on Sherpa and Iris during this free time.
Registration to this session is optional. We would like to see how many people plan to join, and see what operating systems attendees have on their laptops so that we can better prepare for the session.
If you plan on using a Windows machine, please register for the event so that we can provide you Windows instructions for Sherpa.
Register for the event.
Have any questions about the session? Would you like more information on the software tools to be used? Send us an message at modelingws229aas at cfa.harvard.edu
We will be available throughout the week at the Chandra booth in the Exhibit Hall (Booth #815) to answer questions, show demos, and discuss current modeling techniques.
Chair: Giuseppina (Pepi) Fabbiano
- Jamie Budynkiewicz
- Raffaele D'Abrusco
- Janet Evans
- Omar Laurino
- Aneta Siemiginowska
The session will begin with an introduction to the challenges of data analysis, followed by presentations and demos of Sherpa and Iris. Afterwards, attendees will have around a half hour of free time to play around with the tools by themselves or with help from the instructors.
|9:30||Introduction: the challenges of and tools for N-D modeling of complex data in astronomy|
|9:35||Sherpa presentation and demonstration:|
|10:20||Iris presentation and demonstration:|
|11:00||Free time with Sherpa and Iris: work through provided science threads, ask session leaders questions, work through your own science threads|
As it is not uncommon for internet connections at AAS to be slow, we highly encourage users to download Sherpa and Iris before coming to the session.
Note: Iris only runs on Linux or OS X operating systems.
Pre-made science threads for users to work through during the free time part of the session. Thumb drives with the worksheets and datasets will be provided at the meeting for those who are unable to download the files before or during the meeting.
The Sherpa worksheets are Jupyter notebooks which will be run on a server for your convenience during the meeting. If the internet connection is slow, you will be able to download the notebooks and datasets from thumb drives during the meeting.
Sherpa Jupyter notebooks
Browse/download the PDF worksheets and datasets from Github
$ cd <some_directory> $ git clone https://github.com/ChandraCXC/aas229iris $ cd aas229iris $ ls README.md ots/ worksheets/ # if you downloaded the git repository before, pull up latest changes with: $ git pull origin master
Or, download a tar package: aas229iris.tar.gz
Here are the presentation slides for the session. Look at the worksheets to run through the demos.
- Introduction: multi-wavelength research with Sherpa - Pepi Fabbiano
- Sherpa overview and examples (ODP format) - Omar Laurino
- Iris overview and examples - Jamie Budynkiewicz