تقرير
Undercoverage in high-statistics counting experiments with finite MC samples
| العنوان: | Undercoverage in high-statistics counting experiments with finite MC samples |
|---|---|
| المؤلفون: | Alexe, Cristina-Andreea, Bendavid, Joshua L., Bianchini, Lorenzo, Bruschini, Davide |
| سنة النشر: | 2024 |
| المجموعة: | High Energy Physics - Experiment Physics (Other) |
| مصطلحات موضوعية: | Physics - Data Analysis, Statistics and Probability, High Energy Physics - Experiment |
| الوصف: | We consider the problem of setting a confidence interval on a parameter of interest from a high-statistics counting experiment in the presence of systematic uncertainties modeled as unconstrained nuisance parameters. We use the profile-likelihood test statistic in the asymptotic limit for confidence interval setting and focus on the case where the likelihood function is derived from a finite sample of Monte Carlo simulated events. We prove as a general result that statistical uncertainties in the Monte Carlo sample affect the coverage of the confidence interval always in the same direction, namely they lead to a systematic undercoverage of the interval. We argue that such spurious effects might not be fully accounted for by statistical methods that are usually adopted in HEP measurements to counteract the effects of finite-size MC samples, such as those based on the Barlow-Beeston likelihood. |
| نوع الوثيقة: | Working Paper |
| الوصول الحر: | http://arxiv.org/abs/2401.10542Test |
| رقم الانضمام: | edsarx.2401.10542 |
| قاعدة البيانات: | arXiv |
| الوصف غير متاح. |