We introduce a new kind of likelihood function based on the sequence of moments of the data distribution. Both binned and unbinned data samples are discussed, and the multivariate case is also derived. Building on this approach we lay out the formalism of shape analysis for signal searches. In addition to moment-based likelihoods, standard likelihoods and approximate statistical tests are provided. Enough material is included to make the paper self-contained from the perspective of shape analysis. We argue that the moment-based likelihoods can advantageously replace unbinned standard likelihoods for the search of non-local signals, by avoiding the step of fitting Monte-Carlo generated distributions. This benefit increases with the number of variables simultaneously analyzed. The moment-based signal search is exemplified and tested in various 1D toy models mimicking typical high-energy signal--background configurations. Moment-based techniques should be particularly appropriate for the searches for effective operators at the LHC.
A General Theory of Goodness of Fit in Likelihood Fits
可以使用分箱数据和未分箱数据来执行对数据的极大似然拟合。两种情况下的似然拟合都只产生拟合量,而不产生拟合的拟合优度。对于分组数据,在执行极大似然拟合之后,可以通过使用$χ^2$方法获得拟合优度的度量。对于未分组的数据,目前获得了拟合参数,但没有可用的拟合优度测量。迄今为止,这仍然是统计学上一个未解决的问题。通过考虑似然函数关于变量变化的变换性质,我们得出理论预测的概率密度与{ \it the data density}的似然比在变量变化时不变,并提供了拟合优度。我们将展示如何应用此似然比装箱以及未装箱的可能性,并表明,即使是$χ^2$测试是这个一般理论的一个特例。为了计算拟合量的误差,我们需要解决逆概率问题。我们使用贝叶斯定理来做到这一点,使用拟合优度中获得的数据密度。这允许人们在不使用贝叶斯先验的情况下反转概率。由此产生的统计数据与频率论的观点是一致的。
How good are your fits? Unbinned multivariate goodness-of-fit tests in high energy physics
多元分析在高能物理中起着重要的作用。这种分析通常涉及执行概率密度函数(p.d.f.)to the data数据.本文探讨了各种未分组的方法来确定的拟合优度的p.d. f。to the data数据.每种方法的应用和性能进行了讨论的背景下,在现实生活中的高能物理分析(达利兹图分析)。本文提出的几种方法也可用于非参数确定两个样品是否来自同一亲本p.d.f.。这可以用于,例如,确定探测器蒙特卡罗模拟的质量,而不需要效率的参数表达式。
Center for Nuclear Theory, Department of Physics and Astronomy, Stony Brook University(核理论中心,物理与天文学系,石溪大学);Massachusetts Institute of Technology(麻省理工学院);Université Paris-Saclay, CNRS/IN2P3, IJCLab(巴黎-萨克雷大学,CNRS/IN2P3,IJCLab);Institute for Particle Physics Phenomenology(粒子物理学现象研究所);Department of Physics, Durham University(物理系,杜ham大学)
The Well-Tempered Likelihood: Honest Confidence Intervals for Misspecified Models
Stanford University(斯坦福大学);SLAC National Accelerator Laboratory(史丹福直线加速器中心);Massachusetts Institute of Technology(麻省理工学院);Institut des Hautes Études Scientifiques(高等科学研究所);CEA Paris-Saclay(巴黎萨克雷原子能委员会)