So on the seventh day he
rested from all his work drank too much coffee and programmed for 10 hours
Last week I took a one week “break” from sabbatical to travel to New York City for the second annual Astro Hack Week. The conference (or unconference, as I should probably call it) was informal, intense, and great fun. This article on Space.com gives a good summary of the aims of the event. It’s uncommon to attend a conference and really feel a sense of being surrounded by kindred spirits, but this was one of those occasions.
One highlight among many was the Thursday evening “debate” about model selection, which pitched David Hogg against Phil Marshall and myself. The debate was made more interesting by the presence of wine, beer, and snacks. Phil and I were arguing for fairly purist “Bayesian model averaging” ideas as those most aligned with the goals of scientists. Hogg disputed this and argued for an interesting mix of eclectic pragmatism and decision theory. I don’t know if it was clear at the time, but I don’t think we really disagree much at all. Hogg advanced his argument through an example of deciding where to drill holes in a metal plate (experimental design of a fibre spectrograph if I recall correctly), where concrete decisions need to be made in the presence of uncertainty. Here, the decisions being made are really of the form “let’s do X” and not of the form “I don’t know fore sure whether theory A or B is true but if I have to guess, I’ll guess B”. In the latter case Phil and I argued (of course) that posterior probabilities of hypotheses are what you want. Here’s a GIF of a memorable moment from the event, Hogg giving a reductio-ad-absurdum of maximum likelihood, drawing an insane curve that perfectly fits all data points (image credit: Daniela Huppenkothen, who did an amazing job organising everything, and still had the energy to talk new projects after dinner at the end of the week!).
Another great thing about the meeting was getting some things implemented that I’d been putting off for ages, by pair coding with PhD students Alex Malz (NYU), Daisy Leung (Cornell), and João Faria (Porto) on various astrostatistical projects. One great advantage of having other people looking over your shoulder is the improvement to discipline. On more than one occasion I was ready to call it a day and relax, but was nudged to continue.
All in all, a great and productive week. Now I need a holiday.