An Introduction to Error Analysis: The Study of Uncertainties in Physical Measurements

Citation

Taylor, J. R. (2022). An Introduction to Error Analysis: The Study of Uncertainties in Physical Measurements (3rd ed.). University Science Books.

Chapter Summary

The third edition has thirteen chapters, appendices (including the normal error integral and probabilities for chi squared), a bibliography and an index. The publisher's description says the book introduces the study of uncertainties to lower-division science students using familiar examples. Chapter titles below follow the publisher's and catalogue listings, and the descriptions stay within what those titles and the publisher's description support.

Chapters 1–3: Preliminary Description of Error Analysis; How to Report and Use Uncertainties; Propagation of Uncertainties

The opening chapters present the vocabulary and the working rules: what error analysis is for, how to state a measured result with its uncertainty and compare it with other results, and how uncertainties in measured quantities carry through to quantities calculated from them.

Chapters 4–5: Statistical Analysis of Random Uncertainties; The Normal Distribution

The book then moves from rules to statistics. Chapter 4 treats the statistical analysis of repeated measurements, and Chapter 5 presents the normal distribution, the model behind most of the quoted uncertainties.

Chapter 6: Rejection of Data

A chapter on when, and by what rule, a measurement may be discarded. Chauvenet's criterion is among the terms indexed in the book.

Chapters 7–9: Weighted Averages; Least-Squares Fitting; Covariance and Correlation

These chapters deal with combining measurements of different precision, fitting a line or curve to data by least squares, and describing the relationship between two measured quantities.

Chapters 10–12: The Binomial Distribution; The Poisson Distribution; The Chi-Squared Test for a Distribution

The final classical chapters introduce two further probability distributions and a chi-squared test for how well data agree with an expected distribution.

Chapter 13: Bayesian Statistics

The chapter new in this edition. The publisher says it is a full chapter on Bayesian thinking that introduces conditional probabilities and Bayes' theorem, using deliberately simple examples as a first step.

Key Concepts

Chapters 1–3: Reporting and propagating uncertainties

• Uncertainty and best estimate: the pairing in which results are stated.
• Propagation of uncertainties: the rules of Chapter 3.
• Fractional uncertainty: a term indexed in the book.

Chapters 4–9: Random errors and fitting

• Normal distribution: Chapter 5.
• Weighted average: Chapter 7.
• Least-squares fitting: Chapter 8.
• Covariance and correlation: Chapter 9.

Chapters 10–13: Distributions and inference

• Binomial and Poisson distributions: Chapters 10 and 11.
• Chi-squared test: Chapter 12.
• Bayes' theorem: Chapter 13.

Critical Analysis

Strengths

• Clear sequence from practice to theory. The chapters begin with how to report, use and propagate uncertainties and only then develop the statistics behind the rules, which suits laboratory students who need the rules immediately.
• Updated for computing. The publisher says the edition adds chapter-ending problems meant to be solved with computers; almost all are stated for common spreadsheet programs such as Excel, which the publisher argues teaches the meaning of standard deviations, correlation coefficients and normal distributions if students build their own spreadsheets.
• A Bayesian chapter. The new Chapter 13 answers the growth in popularity of Bayesian statistics noted by the author, with simple examples and a limited aim.
• Long track record. The publisher describes the book as a best-selling text for forty years, translated into many languages.

Limitations

• A single, late Bayesian chapter. The Bayesian material is confined to Chapter 13, so a course built around Bayesian data analysis will need further sources.
• Spreadsheet-oriented computing. The new problems are written mainly for spreadsheets; readers who want programming-based analysis will need to adapt them or use other books.
• Limited independent evaluation. No journal review of the third edition was located, so claims about teaching effectiveness rest on the publisher's description.

Where It Fits in the Course

Taylor's book is the measurement and statistics companion in the computational physics list, rather than a numerical methods text. Newman's Computational Physics and Landau, Páez and Bordeianu's Computational Physics: Problem Solving with Python teach algorithms and programming through physics problems; Thijssen's Computational Physics is a graduate-level account of research methods; and Press and colleagues' Numerical Recipes is a reference of algorithms with working routines. Kinder and Nelson's A Student's Guide to Python for Physical Modeling teaches the language.

None of these is devoted to experimental uncertainty, which is Taylor's subject. It is best read alongside a first laboratory course, so that each rule can be applied to the student's own data straight away, and before or alongside the computational texts when their problems ask for fits, error bars or goodness-of-fit.

How to Use This Book

The page metadata says algebra and the beginnings of calculus suffice for the first half, with a little more calculus later. The book is aimed at students in first laboratory courses.

A practical plan:

• Core for any laboratory course: Chapters 1–5, which give the rules for reporting and propagating uncertainties and the statistics behind them.
• Next: Chapters 7–9 when fitting data, and Chapter 6 when deciding about outliers.
• Probability and testing: Chapters 10–12, then Chapter 13 for a first look at the Bayesian approach.
• Order of study: Because later chapters use the statistical vocabulary of Chapters 4 and 5, readers pressed for time should not skip them to reach fitting; Chapters 1–3 alone give rules but not their justification.
• Practice: Do the chapter-ending problems, and build spreadsheets from scratch for the computational ones rather than relying on built-in functions, as the publisher recommends. A separate Student Solutions manual accompanies the third edition.