Precision Standard Model: Observables and Inference
Precision Standard Model work is reliable only when the object being inferred is fixed before the statistic is chosen. A complex pole belongs to an amplitude; a pseudo-observable belongs to a stated pole and radiation convention; a fiducial cross section belongs to a measurement function; and a fit belongs to an exact data release, covariance or likelihood, and nuisance model. This chapter builds that chain from definition to combination. It teaches durable methods, not a live summary of measurements or preferred parameter regions.
Choose the layer before the statistic
Section titled “Choose the layer before the statistic”Several quantities called “an observable” live at different layers. They are related, but they are not interchangeable.
| Layer | Mathematical object | Question it answers | Essential check |
|---|---|---|---|
| Scattering amplitude | Pole and Laurent coefficients of an amplitude between stable states | What resonance information is invariant under field and gauge conventions? | The pole and the recombined amplitude are gauge independent. |
| Pseudo-observable | Pole mass, width, residue-derived coupling, or deconvolved line-shape parameter under an explicit convention | Which compact parameters summarize a resonant amplitude? | The extraction map and removed radiation or background terms are stated. |
| Fiducial observable | Cross section or distribution defined by a particle-level measurement function | What is measured in a reproducible phase-space region? | Object definitions, cuts, bin edges, units, and infrared-safety conditions are complete. |
| Released inference object | Data vector with covariance, or a statistical model with auxiliary data and nuisances | What information may be reused in another inference? | Version, ordering, correlation semantics, corrections, and overlap are known. |
| Combined fit | Joint likelihood or covariance model for shared parameters | What does a specified collection of inputs imply together? | Shared nuisances and duplicate events or constraints are treated once. |
The interfaces are directional. Detector-level counts can be forward-folded from a fiducial or truth-level prediction through a response model. A pseudo-observable can be extracted from a fiducial result only after specifying acceptance, radiation, background, and pole assumptions. A global fit cannot repair an ambiguous observable definition or an incomplete release.
Throughout the chapter, a local pole convention is written explicitly, likelihoods distinguish observed from auxiliary data, and covariance matrices carry an ordering and units. The site-wide metric and natural-unit conventions continue to apply.
Readiness diagnostic
Section titled “Readiness diagnostic”This is an unscored diagnostic. A “not yet” answer points to the shortest repair; it is not a reason to skip the scientific question.
| Can you do this without guessing? | If not, repair it here |
|---|---|
| Expand an amplitude about a resonance pole and explain why the unstable particle is not an LSZ external state. | Review unstable-particle observables and resonance approximations. |
| Translate an electroweak prediction when the chosen input set changes. | Review electroweak renormalization and input schemes. |
| Propagate correlated experimental, parametric, and theory uncertainties without adding the same source twice. | Review validation and theory uncertainties. |
| Separate Higgs production, decay, total-width, and acceptance assumptions. | Review Higgs interactions, production, decay, and pole observables. |
| State the bins, response, nuisance constraints, and overlap needed to reuse a collider result. | Work through the exact synthetic covariance and nuisance model, then return to the release interface below. |
Hard prerequisites on an individual page remain hard: they supply a definition or method used without rederivation. The sequence below is only a suggested reading order. A reader who already has the required background may enter at the layer matching the task.
The five-page route
Section titled “The five-page route”- Standard Model pseudo-observables and unstable particles fixes the amplitude, complex-pole, narrow-width, and production–decay language. If a resonance mass, width, or partial width is being treated as an ordinary stable-particle matrix element, repair the unstable-particle background and start here.
- Electroweak precision observables organizes pole, charged-current, and low-energy quantities in an input scheme with radiative corrections and correlated uncertainties. If a prediction changes when input parameters are merely renamed, repair the input-scheme background first.
- Higgs precision and coupling inference derives rate and differential interfaces, exposes the total-width flat direction, and separates the shorthand from EFT interpretations. If production, branching fractions, and acceptance have been collapsed into one number, repair the Higgs production-and-decay background first.
- Collider measurements, fiducial predictions, and likelihood provenance connects particle-level definitions, detector response, released statistical models, and immutable identifiers. If covariance ordering, nuisance meanings, or dataset overlap are unknown, repair the uncertainty background and reconstruct the release record before reuse.
- Correlated Standard Model fits and consistency tests builds joint likelihoods, generalized least squares, profiling, goodness-of-fit, and stability tests. If inputs share events or nuisance constraints that cannot be identified, stop at the release interface rather than multiplying likelihoods.
One chain, with explicit handoffs
Section titled “One chain, with explicit handoffs”A reusable analysis can be summarized as a sequence of contracts:
| Stage | Minimum information passed forward | What the next stage must not silently change |
|---|---|---|
| Hard-scattering calculation | External stable states, perturbative order, parameter and renormalization schemes | Pole definition or coupling normalization |
| Shower, hadronization, and detector response | Generator and response configuration, object mapping, validation domain | Particle-level observable definition |
| Reconstruction and selection | Dataset identity, object calibration, event selection, binning | Event or bin membership after looking at fit residuals |
| Pole or fiducial interpretation | Pole expansion or measurement function, radiation treatment, acceptance map | Background/interference terms or validity domain |
| Public result | Observations, auxiliary data, covariance or likelihood, exact version and corrections | Nuisance constraints, bin ordering, units, or supersession state |
| Theory comparison | Prediction version, parameter scheme, uncertainty sources and correlations | Correlation model chosen after seeing the preferred result |
| Combined inference | Joint likelihood, overlap resolution, shared nuisances, frozen masks | Dataset versions or trial family during inference |
This chain does not claim that showering, detector simulation, or reconstruction is performed in this chapter. It specifies what those stages must communicate so that the observable and inference remain reproducible.
Informal synthesis check
Section titled “Informal synthesis check”A sound chapter-level work product should meet all of these criteria:
- name the layer of every quoted quantity and write its defining equation or measurement function;
- state the pole, input, renormalization, and parameter conventions needed to compare predictions;
- bind every numerical input to an exact release, bin ordering, units, and correction history;
- encode correlations through a positive-semidefinite covariance or shared nuisance model, without duplicating auxiliary constraints;
- freeze overlap rules and validity masks before fitting; and
- report the test statistic, degrees-of-freedom or calibration procedure, and local-versus-global interpretation needed for the claim.
If the first two criteria fail, return to the pole and sector-specific pages. If release identity or correlations fail, return to the collider interface. If only combination or calibration fails, repair the combined-fit workflow. A result is not made more precise by proceeding through an undefined interface.
As a compact self-check, imagine receiving a two-bin spectrum, a plot of its correlation matrix, and a best-fit coupling. The answer is not reusable yet: the bin values and units, machine-readable covariance or likelihood, nuisance semantics, exact release version, model validity, and overlap with other inputs are missing. The repair is to obtain or construct those objects from the authoritative release—not to digitize the plot and infer a likelihood.
Where to leave the chapter
Section titled “Where to leave the chapter”- To define or calculate a resonance quantity: continue with unstable-particle observables and resonance approximations or unstable-particle effective theory.
- To improve electroweak or Higgs predictions: continue to the electroweak-theory and Higgs chapter.
- To build a collider prediction and its uncertainties: continue to perturbative QCD and partons and validation and theory uncertainties.
- To rehearse covariance and nuisance calculations: reproduce the exact synthetic fixture before using released inputs.
- To interpret deviations through an effective theory: continue to SMEFT, HEFT, and Standard Model observables and the EFT and Standard Model tests research field.