Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Data analysis techniques for physical scientists
Pruneau C., Cambridge University Press, New York, NY, 2017. 716 pp. Type: Book (978-1-108416-78-8)
Date Reviewed: Feb 22 2019

To quote Shakespeare, “What’s in a name?” For Pruneau’s book Data analysis techniques for physical scientists, quite a bit it turns out. This great magnum opus (more than 700 pages) should be required reading for its intended audience--at least if it were titled something like Data analysis techniques for particle physicists. Indeed, after a short introduction to the foundations of the scientific method (13 pages), almost all the examples in both the first (370 pages) and second (350 pages) parts are related, in one way or another, to high-energy physics and the analysis of particle collisions or decays in accelerators such as Brookhaven’s Relativistic Heavy Ion Collider (RHIC) and the European Organization for Nuclear Research (CERN) Large Hadron Collider (LHC). The book ends with a very short third part (43 pages) that focuses on simulation techniques, including a chapter dedicated to the modeling of collisions and detectors.

The first part, “Foundation in Probability and Statistics,” has six chapters. The material is organized in a very structured and gradual fashion. Chapter 2--since chapter 1 serves as the book’s introduction--introduces the key distinctions between the two main interpretations of probabilities, namely the frequentist approach (which assumes an infinite number of similar measurements of a given experiment) and the Bayesian one (that builds upon an incremental refinement of someone’s beliefs about a hypothesis, based on experimental data). Chapter 3 covers the analytical parametrized models of some important probabilistic processes; oddly, the Landau distribution, used to model the energy loss of particles traveling in various media, is only detailed in chapter 8. The basic notions of data population, sampling, statistics, and estimators of the parameters of probabilistic models are detailed in chapter 4, while a more advanced section on Fisher information introduces the idea of measuring the information content of data and models (another approach, based on entropy, is detailed in chapter 7). The two main frequentist approaches to parameter estimation, that is, maximum likelihood and least squares, are described in chapter 5; Kalman filtering, a technique usually addressed in online signal processing textbooks, is also introduced in a very pedagogical manner. Chapter 6 completes this presentation of the frequentist approach in the experimental sciences by introducing the ubiquitous notions of confidence intervals, significance levels (or p-value, oddly missing from the index), and tests such as χ2 and t-test. The 100-page chapter 7 is entirely dedicated to the Bayesian approach and digs into key issues such as the proper choice of priors, inference in the presence of Gaussian noise, estimation in nonlinear models, and model comparison techniques such as the odds ratio.

The second part of the book, “Measurement Techniques,” is a use case in high-energy particle physics for the material presented in the first part. For readers already somewhat versed in data analysis, it can in fact be read independently. Chapter 8 provides a tutorial on the observables and experimental apparatus used in particle decay and collision processes. Chapter 9 then focuses on automatic event, vertex, and track reconstructions (where Kalman filtering shines) in particle detectors, such as those in the LHC. The issue of data analysis based on correlation (and cumulant) functions is the subject of chapters 10 and 11; these functions relate the physical parameters of a particle or group of particles, such as rapidity or transversal momentum, with those of another particle (or group) created by collisions. This data is of significant value at RHIC and LHC when dealing with hadrons, be they protons or heavy ions. In 75 pages, this rather specialized area, which often provides strong evidence for the validity of existing physical models, is covered in detail. The final chapter of Part 2 is dedicated to data correction techniques, or unfolding, in the presence of experimental errors.

The last part of the book, “Simulation Techniques,” looks more like a pair of appendices. Chapter 13 discusses Monte Carlo techniques, which are at the core of current software simulation tools and libraries in high-energy physics, followed by the modeling of collisions and detectors in chapter 14.

Overall, despite some minor editorial kinks and a few typographical errors, I found this book of great value, and I highly recommend it to interested readers, both at the undergraduate and graduate levels. The writing is crisp and pedagogical, and even if a new title would provide more guidance to readers as to what the book is really about, the material provided in the first part remains of great value to all physicists--as good and even sometimes more in-depth than Cowan’s classic textbook [1]. For particle physicists, the second part provides a great overview of the current data analysis techniques, including correlation functions, widely used by the high-energy physics community.

More reviews about this item: Amazon

Reviewer:  P. Jouvelot Review #: CR146443 (1905-0166)
1) Cowan, G. Statistical data analysis. Oxford University Press, New York, NY, 1998.
Bookmark and Share
 
Physics (J.2 ... )
 
 
General (H.2.0 )
 
 
Physical Sciences And Engineering (J.2 )
 
 
Probability And Statistics (G.3 )
 
Would you recommend this review?
yes
no
Other reviews under "Physics": Date
Computational astrophysics
Arnett W. Communications of the ACM 28(4): 354-357, 1985. Type: Article
Sep 1 1985
Computer simulation methods: in theoretical physics
Heermann D., Springer-Verlag New York, Inc., New York, NY, 1986. Type: Book (9780378169660)
Jul 1 1987
Computing in high energy physics
Hertzberger L., Hoogland W. (ed)  Computing in high energy physics,Amsterdam, The Netherlands,1986. Type: Whole Proceedings
Mar 1 1989
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy