The AstroStat Slog

The idea that some useful materials related to the Chandra calibration problem, which CHASC is putting an effort to, could be found from PHYSTAT conferences came along. Owing to the recent advanced technologies adopted by physicists (I haven’t seen any statistical conference offers what I obtained from PHYSTAT-LHC 2007), I had a chance to go through some video files from PHYSTAT-LHC 2007. The files are the recorded lectures and lecture notes. They are available from PHYSTAT-LHC 2007 Program.

PHYSTAT seems to occur biannually and participants are many high energy physicists and a few statisticians. Strangely, those small number of statisticians are extremely renown. This year’s speakers from statistics include Sir David Cox, Nancy Reid, and James Berger. I was more surprised from previous years’ PHYSTAT participants from the statistical community.

The main issue of the past few PHYSTAT meetings, if I were asked to summarize, would be

H_o: no signal vs. H_1: signal (Higgs exist!)

Or in a bit wordy way, it would be H_o: background only model vs. H_1: background and new particle. Due to sophisticated settings of physical experiments (a link to Large Hardron Collider (LHC) from wikipedia; it hasn’t produced data yet but the discovery of the new particle, called Higgs is expected to be happened at this facility), this simple hypothesis testing become tremendously challenging. Knowing background prevents false discovery at the physically nominal level and leads to discover the new particle with a certain level of statistical assertion. The main statistical topics for separating signal from background were multiple testings (due to multiple channels), and coverages and upper limits on background. Both frequentist and Bayesian methods were applied and thoroughly analyzed.

Among all fascinating talks and discussions, I particularly fancy the conversation between Nancy Reid and Sir David Cox as discussants, and audience as questioners (This discussion happened within Radford Neal’s presentation video file. Note that Neal’s talk is hardly audible, which gave me a great disappointment since his topic was of the personal utmost interest. If you want to listen the conversation from Reid, please skip this video file about 45 minutes). In the discussion, Nancy Reid clarified some statistical notions like p-values, coverage or upper/lower limits, power, powerful tests, which physicists tend to overlook. Also a comment from a physicist who criticized the misconception of the confidence interval among physicists made my ears perfect receivers. Above all, the most witty comment I heard was from Loius Lyons after Kyle Cranmer’s talk:

This afternoon session will be at Building Six. … You should allow about 12 plus or minus 3 minutes to get there. That’s statistical error. Don’t allow systematic error, getting lost on the way …

All talks are worthwhile to go through to understand the phystatistical challenges in high energy physics. Personally speaking, physicists, Kyle Cranmer, Luc Demortier, Eilam Gross, Jim Linnemann, and Wade Fisher offered nice summaries particularly for statisticians, who would like to participate in discovering the new particle.