wunder beta

📘 How do you report error in physics?

Propagation, distributions, and honest ±—how 400-level physics turns a measurement into a defendable result.

14
lessons
~30 min
to learn
Adults
level
Start the course →

What you’ll learn

  1. Foundations of Measurement UncertaintyDistinguish Type A and Type B evaluations and map them onto the GUM framework used in modern physics.The chapter establishes that uncertainty is a parameter characterizing the dispersion of values. It separates statistical evaluation from other means and shows why both enter every physics result. This distinction structures all subsequent chapters.
  2. Random versus Systematic ComponentsDecompose an uncertainty budget into random and systematic contributions and propagate their distinct statistical properties.Random components average down with repeated measurements while systematics remain fixed. The chapter demonstrates how to identify, quantify, and combine both classes without double-counting.
  3. Probability Distributions in MeasurementSelect and justify the appropriate distribution (normal, Student-t, Poisson, or uniform) for a given experimental observable.The chapter reviews how finite sample size, detector dead time, and background rates dictate distribution choice. It shows the numerical consequences of using the wrong model on coverage intervals.
  4. Law of Propagation of UncertaintyApply the full multivariate propagation formula including all first-order partial derivatives and covariances.The chapter derives the general expression for variance of a function of several variables and demonstrates its reduction when variables are uncorrelated. Worked numerical examples from atomic physics illustrate the effect of neglected covariances.
  5. Covariance Matrices in Multi-Parameter FitsConstruct, interpret, and diagonalize experimental covariance matrices arising from shared calibration or background models.The chapter shows how shared systematics generate non-zero covariances and how eigenvalue decomposition reveals the effective number of independent constraints. It quantifies the bias introduced by treating correlated data as independent.
  6. Least-Squares and χ² MinimizationPerform weighted least-squares fits, extract parameter uncertainties from the Hessian, and test goodness-of-fit with the χ² distribution.The chapter derives the normal equations, shows how the covariance matrix is obtained from the second-derivative matrix, and explains when χ² probabilities indicate model failure rather than statistical fluctuation.
  7. Maximum-Likelihood EstimationWrite the likelihood function for binned and unbinned data and extract both point estimates and profile-likelihood intervals.The chapter contrasts likelihood with least-squares, demonstrates construction for Poisson and Gaussian cases, and shows how to obtain proper coverage when the likelihood is non-parabolic.
  8. Bayesian Methods for Systematic UncertaintiesConstruct priors for systematic uncertainties, sample the joint posterior, and report credible intervals that incorporate both statistical and systematic contributions.The chapter introduces Markov-chain Monte Carlo sampling for high-dimensional posteriors and shows how to choose weakly informative priors that do not dominate the data. Coverage properties are compared with frequentist intervals.
  9. Monte Carlo Techniques for Complex Error PropagationDesign and validate Monte Carlo error propagation for non-linear, non-Gaussian, or cut-dependent observables.The chapter explains how to sample input distributions, apply the full analysis chain, and extract quantiles. It addresses convergence diagnostics and the computational cost of rare-event tails.
  10. Systematic Uncertainties in Detector CalibrationIdentify, quantify, and decorrelate calibration systematics that affect an entire data-taking period.The chapter walks through in-situ calibration strategies, shows how control samples constrain nuisance parameters, and demonstrates the construction of a reduced covariance matrix after marginalization.
  11. Blind Analysis and Analysis ChoicesImplement blinding protocols and quantify the impact of analysis decisions on final uncertainties.The chapter reviews blinding techniques used in precision experiments, shows how to preserve coverage while allowing data-driven corrections, and discusses the statistical price of multiple testing.
  12. Look-Elsewhere Effect and Trial FactorsCalculate trial factors and convert local to global significance for searches over continuous or discrete parameter spaces.The chapter derives the asymptotic formula for the look-elsewhere effect, demonstrates its application to resonance searches, and shows how to validate the correction with pseudo-experiments.
  13. Reproducibility and Uncertainty Reporting StandardsProduce an uncertainty budget that satisfies current community standards for reproducibility and meta-analysis.The chapter reviews PDG and GUM reporting requirements, shows how to document nuisance parameters, and explains why incomplete budgets prevent legitimate combination of results.
  14. Case Study: Gravitational-Wave Parameter EstimationApply the full error-analysis toolkit to a modern gravitational-wave event and produce a publication-ready uncertainty statement.The chapter integrates propagation, covariance, likelihood, and systematic marginalization using the actual LIGO analysis chain. It demonstrates how the final credible intervals change when each layer is added sequentially.

Questions this course answers

A calibration certificate states that a 10 V reference has a 0.5 ppm uncertainty at k=2. In the GUM budget this contribution is entered as which quantity?

The certificate supplies information obtained outside the present statistical data set, so the component is Type B. Its standard uncertainty is obtained by dividing the stated expanded uncertainty by the coverage factor k=2, yielding u_B = 0.25 ppm.

A new frequency comb measurement yields 200 repeated readings whose standard deviation of the mean is 3 Hz. The manufacturer quotes a 5 Hz uncertainty (k=1) on the reference oscillator used to stabilize the comb. Write one sentence that states the combined standard uncertainty and justifies why both numbers are required.

The GUM requires that every source of uncertainty be expressed as a standard uncertainty and combined in quadrature regardless of whether it was obtained statistically or by other means. Omitting the Type B term would understate the dispersion that must be reported to users of the frequency value.

In a dataset of 400 independent runs the statistical uncertainty on the mean is 0.25 %. The dominant systematic uncertainty of 1.8 % arises from a single shared calibration constant. What is the combined standard uncertainty after all runs are averaged?

The random term scales to 0.0125 % while the 1.8 % systematic remains fixed; adding in quadrature gives √(0.0125² + 1.8²) ≈ 1.82 %.

Place the following steps in the order required to propagate a mixed random-plus-systematic uncertainty budget.

Classification must precede scaling; the covariance matrix is assembled only after the scaled random block and the fixed systematic block are known; the final quadrature sum is performed last.

An ion-trap experiment records 7 detected photons in a 10 ms window. Which distribution supplies the correct 68 % coverage interval for the underlying rate?

Photon counts are discrete rare events; the Poisson model enforces variance = mean and supplies asymmetric intervals that correctly reflect the lower tail at small λ.

A calibration run yields only eight voltage readings whose sample standard deviation is used to quote the uncertainty. Explain why quoting ±s/√8 with the normal quantile produces an interval whose actual coverage is below the nominal 68 %.

With N=8 the estimated standard error is itself a random variable; its ratio to the true σ follows Student's t with 7 degrees of freedom, widening the correct interval by the factor t_{0.68}(7)≈1.11.

Grounded in trusted sources

  • OpenStax
  • NIST
  • Particle Data Group
  • Taylor — An Introduction to Error Analysis
  • NIST — Guidelines for Evaluating and Expressing the Uncertainty of NIST Measurement Results (TN 1297)
  • Particle Data Group — Review of Particle Physics, statistics reviews

Every Wunder lesson is built from real, reputable sources — never invented.

Related courses

📘How do you write a physics lab report?📘How do physicists compute with machines?📘How do you actually read a physics paper?📘Keep proposal from meaning literature dump📘Keep question from meaning trivia prompt📘Keep experiment from meaning gear flex📘Keep analysis from meaning symbol soup📘Visualization: Physics Capstone

Wunder is a personalized learn-anything platform — tell it any topic and it builds a beautiful, fact-checked course in minutes, with narration, a knowledge check, and a college-style University track.

All topics · Home

© 2026 Wunder Learning LLC · Terms & Privacy