In the course of Statistics for Data Analysis I give every spring to PhD students in Physics I spend some time discussing the apparently trivial problem of evaluating the significance of an excess of observed events N over expected background B. 

This is a quite common setup in many searches in Physics and Astrophysics: you have some detection apparatus that records the number of phenomena of a specified kind, and you let it run for some time, whereafter you declare that you have observed N of them. If the occurrence of each phenomenon has equal probability and they do not influence one another, that number N is understood to be sampled from a Poisson distribution of mean B. 

About a month ago I was contacted by a colleague who invited me to write a piece on the topic of science outreach for an electronic journal (Ithaca). I was happy to accept, but when I later pondered on what I would have liked to write, I could not help thinking back at a piece on the power and limits of the use of analogies in the explanation of physics, which I wrote 12 years ago as a proceedings paper for a conference themed on physics outreach in Torino. It dawned on me that although 12 years had gone by, my understanding of what constitutes good techniques for engagement of the public and for effective communication of scientific concepts had not widened very significantly. 

At a recent meeting of the board of editors of a journal I am an editor of, it was decided to produce a special issue (to commemorate an important anniversary). As I liked the idea I got carried away a bit, and proposed to write an article for it. 

Muon tomography is one of the most important spinoffs of fundamental research with particle detectors -if not the most important. 

I recently held an accelerated course in "Statistical data analysis for fundamental science" for the Instats site. Within only 15 hours of online lectures (albeit these are full 1-hour blocks, unlike the leaky academic-style hours that last 75% of that) I had to cover not just parameter estimation, hypothesis testing, modeling, and goodness of fit, plus several ancillary concepts of high relevance such as ancillarity (yep), conditioning, the likelihood principle, coverage, and frequentist versus bayesian inference, but an introduction to machine learning! How did I do?