Can statisticians become experimental physicists?

Before you brush off this post with the “sure” response, let me qualify the title. Of course, anyone can become a particle physicist, although the learning curve can be steep and difficult to climb. But what I mean here is that a student who has been trained as a statistician (through his bachelor’s and master’s) can become a successful experimental particle physicist, without investing any other years of his life studying quantum mechanics and many other arcane physics topics?

The question, in my opinion, is important because of the way we do science with these huge particle detectors these days. Living and working within the large scientific collaborations that conducted the CDF experiment and later the CMS experiment, I was fortunate enough to observe a statistically significant sample of PhD students (myself included), almost exclusively trained in fundamental physics, being hit by a truck – in the form of the set of skills necessary to complete a thesis on a subject of experimental physics during a collider experiment.

The work of a Ph.D. student in a HEP experiment relies little on his painstakingly acquired skills in physics. On the contrary, most of their time (if we exclude most of it, which is gobbled up by attending mostly useless meetings, and a few other smaller “service tasks”, which usually don’t require the skills of a qualified physicist) is spent writing computer programs and scripts, submitting batch jobs to computer queuing systems, and performing advanced statistical analyzes on the resulting datasets.

At this point, I’m kicking myself as this post would have been much more informative and objective if I had taken the time to run a simple Twitter poll before I started writing it. I would have phrased it this way: what percentage of time does a typical Ph.D. student of an LHC experiment goes to

A) the activities that most often require high-level knowledge of particle physics
B) activities that most of the time require high-level knowledge of statistics to be followed
C) activities that mainly require other skills?

Still, I think we can continue the discussion here even without the data, if we take my personal estimate: A – 20%; B-50%; C-30%. Yes, that’s my point: I mean a Ph.D. in experimental physics on a particle collider requires more training in statistics than training in physics. Of course, a doctorate. the pupils are intelligent: if they are confronted with a problem, they go to the library and read a book; or they google a solution to this. Thus, the fact that a significant training in statistics is required to carry out tasks relevant to their doctorate. completion of the thesis is not an obstacle for a physicist.

[Incidentally, note that I could have framed the question differently, including the request of an assessment of computing skills requirements; but here I do not need that nuisance parameter to make my point.]

If you accept for a second my estimates above of the type of education required for a Ph.D. student at the LHC, or even if you want to modify them by non-revolutionary amounts, you will conclude like me that it would be good to open the door of a doctorate. in particle physics to young statisticians.

Of course, a student who has earned his master’s degree in statistics is unlikely to be attracted by the low salaries and high job volatility of an academic career as a HEP researcher. But if they are young and dumb, some may fall under the siren song of physics. So what Bruno Scarpa, professor of statistics at the University of Padua, suggested to me is to propose a course of “Particle physics: foundations, instruments and methods of analysisto the Masters in Statistics there.

Being the lazy bum that I am, I dodged his enticing offer for a while, but eventually gave in, under the pressure of my own beliefs mentioned above. So this semester, I’m teaching this course, and surprisingly enough, there are a dozen kind souls who decided to take it. So far, so good.

My outlook is as follows: out of these 12 students, I would be delighted if one or two asked me for a thesis based on the analysis of CMS data. If then one of them applied for a doctorate. in physics next year, I’ll definitely be back here to say the plan worked!

But let’s slow down. Between a statistician and a physicist, there is a language barrier, made up of the different concepts to which they have been exposed during their studies. So my job these days is, more than teaching particle physics, to bridge that gap. I discussed ancient quantum theory, special relativity, wave-particle duality, quantum mechanics, symmetries and conservation laws, the quark model, deep inelastic scattering, etc. In the second part of the course, however, I will try to stimulate their inner statistician by touching on real topics of physics data analysis. We’ll see how it goes!

Tommaso Dorigo is an experimental particle physicist who works for the INFN at the University of Padua, and collaborates with the CMS experience at CERN’s LHC. He coordinates the European network AMVA4NewPhysics as well as accelerator physics research for INFN-Padova, and is editor of the journal Reviews in Physics. In 2016 Dorigo published the book “Anomaly! Collider Physics and the Search for New Phenomena at Fermilab”. You can get a copy of the book on Amazon.

Comments are closed.