Poster abstract details

Testing the Bayesian Approach to ON/OFF Signal Measurement Problem with High Energy Data
Mateusz Janiak and RafaƂ Moderski


Poster presents a numerical implementation of a solution to so-called "On/Off problem" provided by bayesian statistics as presented by Tom Loredo. "On/off" problem is common in high energy physics and astropysics and deals with estimating both the source signal rate and a detection significance based on observations in on- and off-source regions. Typical approach used nowadays relies on frequestists methods combined with sophisticated ways to calculate significance (which definition is not unique) and detection upper limits. This leads to results that lack of physical meaning e.g. a source signal rate which is negative or a positive source detection due to statistical background fluctuation. We present a practical application of methods proposed by Tom Loredo to real data obtained with H.E.S.S. telescopes. We show a comparison between typical methods and bayesian statistics based on simulated and real data and verify the applicability of proposed solutions.