Sunday, 23 October 2016

#GaiaSprint Summary

This past week (October 17th-21st) I had the pleasure of attending the #GaiaSprint in NYC, hosted by the Simons Center, and led by David Hogg. This aim of this meeting was to take the data released in Gaia DR1 and basically see what we could do with it.

The week started with each participant giving a 2 minute pitch, introducing themselves, what they were bringing to the sprint in terms of data, expertise, code etc, and what they hoped to do. Some people came with well established plans that were well underway, while others (like myself) had ideas that were less far down the road of being implimented. All of us came with a plan to collaborate and share ideas with the other 50 people we were spending the week with.

My plan was to use the parallaxes in the TGAS sample to derive the RR Lyrae period luminosity relation in the mid-infrared. This sounds pretty simple, but rather than just take the distance to each star as 1/parallax and fit a straight line to the data using least squares, I wanted to implement a bayesian model that took into account the uncertainties on the parallaxes, extinctions, apparent magnitudes, and basically everything that could possibly go into the PL relation. As we'll be applying this PL relation to lots of different populations (see Beaton et al. 2016 for a summary of the distance scale part of our program) we need to take into account everything that could affect the PL, and how well we can measure those parameters.

It turns out that going from fitting a straight line to (x,y,dy) data, to a full generative model for the PL relation is pretty ambitious. I'd done some work on the project before I arrived so I had some of the machinery already in place, but I really struggled with getting my head around the statistics. With some (a lot) of help from Adrian Price-Whelan, Branimir Sesar, and Kathryn Johnston (who are all collaborators on the Galactic structure part of this program) I was able to get to a point where I have all the likelihoods and priors for my parameters set. Now I just need to finish implementing the thing.

There was a huge amount of work came out of the meeting, all of which is summarized in Friday's wrap up slides. Each of us presented 1 slide on what we'd done, and it was really amazing to sit there on Friday and see what we'd achieved in just a few days.

Working in this environment for the week, with no real schedule for the meeting (breakout sessions were decided in the morning of each day) was very different from any other meeting I have been to. For a working meeting like this, not having a fixed schedule is great - it lets you get on with exactly what you need to be doing without worrying that you need to get to a good place to stop in 5 minutes because there's a talk. It is very different to how I normally work though.

It's taken me a really long time to find a good way to work. Just over a year ago, at the grand old age of 31, I was diagnosed with ADHD. Until that point, I thought that I was just good at procrastinating, bad at reading and being organized, and I didn't think that ADHD was even a thing that adults had. Especially not women. It turns out that I was very, very wrong. Since then, I've been taking medication, I speak with my coach every couple of weeks (or more frequently if I need to), and I use a number of strategies to keep myself on task and stop my brain getting overwhelmed. One of the major strategies is PLANNING.

Each day, I make a plan. I have a plan for what needs doing, what the priorities are, what depends on what else. It's a bit like coding. Until I did this every day was completely unscheduled. I'd show up, think 'what should I do today?' And randomly choose. Part of ADHD is a tendency to hyperfocus, so I could spend 12 hours making a plot look nice, when really it didn't matter, because the plot had been telling me for the past 11 hours that my photometry was wrong. Or I'll get stuck in circles, trying to solve a problem by writing code, when I really should take a walk and come back and write it out on the whiteboard.

This was the only problem I had with the week, and it's my problem. I felt under a lot of pressure to go with the 'unscheduled' nature of the week, even though I know that working that way is completely bad for me. If I don't write on my plan '12:30 go outside' I won't go outside, and if I don't go outside, my brain will lock up. Because I wasn't as productive as I knew I could be (because I now know that when I am working properly, I am fucking good at this), I ended up spending my evenings working until midnight to try to catch up to where I thought I should be. (Imposter syndrome FUCK YEAH!) But I know for next time, and it's an important lesson to learn.