PHYS366: Statistical Methods in Astrophysics

Adam Mantz & Phil Marshall

A more detailed version of this course description can be found here

This is a course about data analysis, with examples from astronomy.

It is designed to be interactive, and useful both to graduate students in the physics department at Stanford and elsewhere around the world.

"Existing and emerging statistical techniques and their application to astronomical surveys and cosmological data analysis. Topics covered will include statistical frameworks (Bayesian inference and frequentist statistics), numerical methods including Markov Chain Monte Carlo, and machine learning applied to classification and regression. Hands on activities based on open-source software in python."

Goals

• Develop familiarity in working with various types of (astronomical) data
• Understand the key role of modeling in data analysis
• Be able to critically evaluate and apply commonly used statistical inference methodologies and advanced statistical reasoning to problems you are likely to encounter in your research.

This means that:

• We will primarily look at examples using image data, and numerical catalogs/databases.

• We will focus on how to think about data analysis as much as how to execute it.

In Class

We will:

• Discuss concepts in data analysis in some depth
• Work through examples together.
• Provide hands-on experience through, collaboratively worked problems

The course is divided into "chunks", which should each take (very roughly) 30-40 minutes to work through.

References

At the beginning of each chunk we'll list some opportunities for additional reading, mostly from:

• David MacKay, "Information Theory, Inference and Learning Algorithms" (free download)
• Ivezic et al, "Statistics, Data Mining and Machine Learning in Astronomy"
• Gelman et al, "Bayesian Data Analysis" (2nd ed.)
• Ross, "A First Course in Probability" (7th ed.)
• Fishman, "A First Course in Monte Carlo"

Grades

1. Class Participation: 20%

2. 8 Weekly Homework Assignments: 45%

3. Final Project: 35%

Projects will be presented during the final class on March 16, and final written reports will be due on March 23.

Homework

• Each homework will consist of a short exercise and a longer problem.

• Assignments will be made available via the 2017 homework repo.

• Submission is by pull request, from your fork to the base repo.

Please make your own folder (e.g. HW1/drphilmarshall) so we can easily identify your solutions.

Presenting Homework

Each student will present (at least) one homework solution in class, in the form of a 5-10 minute talk.

• Collaboration on both solutions and presentations is encouraged.
• Collaborators' input must always be properly cited (e.g. by name, hyperlinked to their solution).

Presenters will be chosen when the assignment is released (if not earlier). More details here.

Projects

• Weeks 1-5: brainstorm ideas, form teams

• Week 6 at the latest: arrange to pitch your idea to Adam and/or Phil, and write a brief abstract

• Weeks 6-10: work on the project, with your teammates.

• March 16: presentations

• March 23: turn in written report by 6PM

Communication

• "Watch" the 2017 homework GitHub repo for messages via its issues system

• Use the Physics 366 Slack team to instant message the instructors and students

Bug one of us if you haven't been invited to join the Slack team by the second week of class, or if you have questions about GitHub.