Session 2WB: New Data Analysis Methods II

Bayesian Analysis and Interpretation of Heavy-Ion Collisions

Heterogenous petascale data sets have been collected at RHIC and at the LHC for heavy-ion collisions. These data are interpreted by commensurately sophisticated multi-component and numerically expensive dynamical models involving numerous unknown parameters. I will show how the model/data comparison is addressed using Bayesian approaches featuring model emulators. In addition to providing a means to rigorously constrain model parameters and make quantitative conclusions concerning the field's most pressing questions, I will show how Bayesian approaches can identify the constraining power of specific classes of observables for determining specific parameters and properties of the novel matter formed in heavy-ion collisions.   

Chisquare Fitting When Overall Normalization is a Fit Parameter

Problems with the use of the $\chi^2$ method to fit an event histogram when the total expected number of events is not fixed, keep appearing in experimental studies. It appeared in our MiniBooNE experiment. It was named Peelle’s Pertinent Puzzle (PPP) in Britain. This puzzle was also found in a NIM article in 1994 by an Italian physicist who has made some major contributions to statistics used by physicists. The puzzle is that in a $\chi^2$ fit, if overall normalization is one of the parameters to be fit, the fitted curve may be seriously low with respect to the data points, sometimes below all of them. This problem and the solution for it are well known within the statistics community, but, apparently, not well known among some of the physics community. The purpose of this talk is didactic, to explain the cause of the problem and the easy and elegant solution.   

Global fits for deep inelastic scattering and related processes

Several decades of high-energy scattering experiments have given us intriguing, though limited, glimpses into the inner structure of protons and neutrons. With the 12 GeV nuclear physics program at Jefferson Lab underway, and plans being made for a future Electron-Ion Collider facility, we are at the threshold of imaging the nucleon’s three-dimensional structure through its quark and gluon quantum probability distributions. Extracting the quantum distributions from the experimental data is very challenging, however, because of the inverse problem: the measured cross sections are given by convolutions of the quantum probability distributions with process-dependent hard coefficients that are perturbatively calculable from Quantum Chromodynamics. While most previous analyses have been based on the maximum likelihood approach, it has become evident that Bayesian likelihood methods are needed, using Monte Carlo sampling techniques to thoroughly explore the parameter space associated with the quantum probability distributions. I this talk I will review recent developments in tackling the inverse problem for extracting quantum distributions from experimental measurements.