What I learn from Astrostatistics Summer School at Penn State

During my first summer as a graduate student at MIT, I got an opportunity to participate in the 12th Astrostatistics Summer School at Pennsylvania State University (Penn State), organized by Eric Feigelson in the Department of Astronomy and Astrophysics. My first impression about the summer school is mostly about some fancy statistical techniques that astronomers are adopting in their research to do data analysis. But the reality is much more than I can ever imagine.

Learn Bayesian Statistics from Thomas Loredo

Instead of diving into different fancy technique in statistics, we started the summer school just like every statistic classes in college start, "Probability." But this time, we got to learn from the head of the Department of Statistics at Penn State, David Hunter. It was a great opportunity to refresh some of these materials after saw it 2-3 years ago. Then, we continued into more general topic in statistics, including "Inference" by Kwame Kankam, "Introduction to Regression" by David Jones from Harvard, and "Model Fitting and Bootstrap" by G. Jogesh Babu. This is the first time that I learn about Bootstrapping after hearing it from different places many times. Even though we do not have time to go over all the details about the technique, I would say that I have a little understanding about the technique and how important it is in astronomy. Instead of assuming underlying distribution for the sample, Bootstrapping essentially randomly draw samples of actual data multiple times to get a set of samples that we can do statistical analysis on. This set of samples have statistics properties similar to sample that we draw from correct distribution.

We also learned about "Multivariate, Clustering & Classification" and "Nonparametric Statistics" from Jessi Cisewski from Yale, Time Series from Eric Feigelson, and "Bayesian Analysis" from Thomas Loredo from Cornell. This is also the first time that I formally learned about clustering and classification outside the context of machine learning. But, I think the main point that I got from this summer school is from Prof. Loredo on Bayesian Statistics. As he explained, most astronomical data analysis is more suitable to Bayesian analysis than Frequentist approach as we often than measurement uncertainty and know some prior knowledge about the problem. Bayesian analysis allowed us to incorporate this information explicitly, which makes the analysis very clear.

Apart from learning about statistic, we also learned a basic of R (a programming language of the statisticians, by the statisticians, and for the statisticians) from Derek Young. Even though I learned a little bit of R from both my statistical classes and my social science classes, it was still a great review and good introduction to different part of the program that I have never heard before, such as nls (non-linear regression), boot (Bootstrapping), princomp (Principle Component Analysis), etc.

R Studio for running R

In addition, we also had a half day field trip to Penn Cave, where we took a boat ride along the all-water limestone cavern. Although it was rather hot outside, the temperature inside the cave is surprisingly cooler. The cave was huge and we saw different type of limestone inside the cave. Plus, we also spotted bison and deers around the lake near the cave.

Penn Cave

In conclusion, the Astrostatistics Summer School at Penn State was a great introduction to different field in statistics that astronomers started to use in their analysis, but it was, by no mean, detailed enough to be useful in any cases. It's up to us to continue learning about this topic and adapt it in our own research in order to make the most from our time here.

PS: an unexpected journey happened the night that I supposed to leave State College. My flight got canceled. It was "canceled", and not just one of those delayed flight. Fortunately, I contacted Noah, my astrophysics friend at Penn State to spend a night. I got a chance to visit his campus, his office, and also a laboratory for making X-ray instrument. I was a quick unexpected journal that was interesting enough and didn't make me feel like I wasted one day in town. Oh, I also went to my first ever Game of Throne viewing party. But what I liked more was the TV show called Bob's Burger!