Online workshop on 2sample/Goodness of Fit tests in Particle Physics
This remote meeting will deal with issues related to Goodness of Fit and with 2-sample testing, especially in more than one dimensions. The talks will be by Particle Physicists and by Statisticians, and will include traditional and also Machine Learning approaches. This is a topic which is relevant to many Particle Physics analyses. It includes applications such as:
- Is the current performance of our detector consistent with established norms?
- Is my data consistent with the Standard Model expectation (or should I suspect New Physics)?
- Are fast simulations consistent with full simulations?
The timetable will allow time for ample discussion.
QUESTIONS
A list of questions of interest:
- Are there situations when conventional GoF outperforms ML approaches?
- Are there possible signals that would show up in multi-D data, but not in lower dimensions?
- Is it possible to adjust the power of ML methods to specific types of alternative hypotheses?
- Do any conventional GoF methods retain power for multi-D data?
- What is a sensible set of benchmark hypotheses for checking power?
- If a discrepancy is discovered, how is it possible to evaluate the LEE effect for ML methods (and for conventional GoF?
- Is it useful to regard NN methods for GoF (e.g. like Wulzer’s) as an extension of the saturated method of Baker and Cousins?
- Are there extra systematics to be included for ML approaches?
PHYSTAT
The PHYSTAT series of Workshops started in 2000. They were the first meetings devoted solely to the statistical issues that occur in analyses in Particle Physics and neighbouring fields.The homepage of PHYSTAT with a list of all workshops and seminars is at https://espace.cern.ch/phystat .
