In the context of searches at the LHC a few generic questions should be
asked, such as: How can we establish if a possible excess of events is
significant? When are we convinced that we have found a real effect? Are
we willing to reject our previous models and hypotheses?
In this seminar I will give an introduction to statistical methods in
particle physics, mainly focusing on frequentist hypothesis testing.
After briefly revisiting some basics I will introduce concepts and the
related jargon of frequentist hypothesis testing. I will proceed to
explain the ingredients of the well-known Brazil plots, taking the Higgs
search and discovery in LHC Run 1 as a historical example. Then, I will
introduce in detail some of the test statistics commonly used by the
large LHC experiments and also mention their modifications in the
Feldman-Cousins and the CLs methods. Finally, I will briefly introduce
the look-elsewhere effect.