A METHOD WHICH ELIMINATES THE DISCRETENESS IN POISSON CONFIDENCE-LIMITS AND LESSENS THE EFFECT OF MOVING CUTS SPECIFICALLY TO ELIMINATE CANDIDATE EVENTS

In setting upper limits in counting experiments, a method exists which removes the discreteness in limits normally present for the Poisson probability distribution. The method, while essentially the same mathematically as a controversial statistical technique proposed around 1950, is now given more physical meaning: event quality information supplements the pure number of candidate events in calculating an upper limit. The method ameliorates the problem in searches far rare processes whereby small changes in data selection criteria (cuts) can have large, discontinuous effects on the limit quoted. It may thus have the potential to ease the reconciliation of different choices of cuts, while reducing the potential for bias in cut selection when one has knowledge of potential candidate events. Paradoxically, the method generally gives more restrictive upper limits than the usual method (assuming no background subtraction in either method), while strictly adhering to the classical definition of confidence intervals. Background subtraction is not yet incorporated.