Search Procedures in High Energy Physics
Luc Demortier · 2009
The usual procedure for searching for new phenomena in high energy physics involves a frequentist hypothesis test followed by the construction of an interval for the parameter of interest. This procedure has a couple of well-known flaws: the eect of the test on subsequent inference is ignored, and in some circumstances the size of the reported interval does not properly reflect experimental conditions. Furthermore, proper treatment of nuisance parameters in a frequentist context is nearly always troublesome. For these reasons there has recently been considerable interest in applying the ideas of Bayesian reference analysis to this problem. We describe ongoing work to calculate reference priors for search experiments, via both numerical and analytical methods. We also show how intrinsic interval estimation can provide a very elegant solution to the testing problem in search procedures.