Astronomy · Observations that support cosmology

From observations to conditional cosmological conclusions

Build an observational handoff that preserves measurement conditions, then test how two distinct assumptions can mimic the same distance inconsistency.

Continue in the approved chapter order. Necessary vocabulary, quantities and assumptions are explained in the lesson.

Back to the chapter

After reading, you can

Distinguish detector data, calibrated observables, summary statistics and fitted model parameters.
Calculate a conditional distance-consistency ratio and two alternative algebraic explanations.
Specify the information needed before combining observations or interpreting a disagreement physically.

A conclusion has an observation history

A recorded count, a calibrated flux, a correlation statistic and a fitted cosmological parameter are different stages of inference. Each can be useful evidence, but they retain different dependencies. Calling all of them data without qualification hides the transformations that make a scientific claim assessable.

The preceding lessons followed those transformations through spectral redshift, calibrated distance indicators, galaxy maps, lensing and background radiation. Their shared lesson is that a physical interpretation needs both an observation and a model of how that observation came to be measured.

An observational handoff should identify the actual measured quantities and their units, reference frames, time windows and response. It should retain calibration, selection, processing choices, uncertainty and covariance, together with quality exclusions and the population to which the result applies.

It should also state which quantities were supplied by a model rather than independently measured. A coordinate inferred from redshift under a distance model cannot silently become an independent distance check of that same model. Reusing information can be valid, but its dependency must remain visible.

Keep the comparison rule with the result

A likelihood describes the probability or probability density assigned to possible data under a stated model and its parameter values. It includes assumptions about measurement noise and often additional nuisance quantities, such as calibration offsets or foreground amplitudes. It is not itself a posterior probability for the model.

Posterior inference adds priors and normalization. A best fit, an uncertainty interval and a model comparison answer different questions and can respond differently to those assumptions. The result needs the model, dataset combination and statistical convention that produced it.

ESA’s Planck PR3 cosmology products illustrate this separation through distinct spectra, likelihood resources, parameter chains and best-fit products. Those products refer to specified data combinations and models. A best-fit parameter is therefore a conditional summary of a comparison, not a direct value written into one map pixel.

Similarly, a supernova distance estimate carries light-curve modelling, passbands, calibration, population and selection conditions. A time-delay lens result carries lens geometry and mass-model freedoms. Combining such results requires the appropriate measurement relationships rather than merely placing quoted numbers in the same column.

State the framework before testing its consistency

For an original arithmetic example, adopt a transparent FLRW distance framework with the relation DL = (1 + z)² DA. Here DL is luminosity distance and DA is angular-diameter distance, using the same source and cosmological redshift z. Transparency and the adopted propagation framework are conditions of this exercise.

Stipulate z = 0.50 and DA = 1000 Mpc as the reference quantities. The relation predicts DL = 1.50² × 1000 = 2250 Mpc. These values are supplied teaching inputs, not an actual observational determination or a fit of an expansion law.

Now suppose a separately supplied flux-based distance proxy is 2500 Mpc under its stated intrinsic-luminosity calibration. Form η = DL,proxy/[(1 + z)² DA] = 2500/2250 = 10/9 ≈ 1.111111111.

For these exact supplied values, η differs from 1. That is an algebraic discrepancy with the full collection of assumptions, not a diagnosis of a new physical effect. If these were measured quantities, uncertainty and covariance would be needed before assigning significance to their difference.

Original conditional consistency exercise, with exact stipulated inputs and no assigned measurement uncertainties.
Quantity	Supplied or calculated value	Role
Cosmological redshift z	0.50	Supplied reference input
Angular-diameter distance DA	1000 Mpc	Supplied reference input
Transparent-framework DL	2250 Mpc	Calculated from the stated relation
Flux-based proxy DL,proxy	2500 Mpc	Separate supplied proxy
Consistency ratio η	10/9 ≈ 1.111111111	Dimensionless diagnostic of the combined assumptions

Two alternatives produce the same proxy

Consider one alternative explanation in isolation: grey attenuation transmits a fraction q = 0.81 of the flux without changing the adopted redshift or angular-distance reference. Retain the correct intrinsic luminosity L and the true reference DL = 2250 Mpc. The received flux is then F = qL/(4πDL²).

If the proxy calculation incorrectly assumes q = 1, it infers DL,proxy = √[L/(4πF)] = DL/√q = 2250/0.90 = 2500 Mpc. Correcting for the independently known transmission would restore √q × 2500 = 2250 Mpc. This is a hypothetical demonstration, not evidence that such attenuation occurred.

As a separate alternative, keep propagation transparent but assume an intrinsic luminosity larger than the true value. Because the inferred distance scales as √Lassumed, the same proxy ratio 10/9 follows from Lassumed/Ltrue = (10/9)² = 100/81 ≈ 1.234567901.

These alternatives are not jointly applied in this exercise. Either one alone can generate the supplied ratio while the reference distance relation remains true. The numerical ratio therefore cannot, by itself, distinguish attenuation from luminosity miscalibration, or exhaust the other possible calibration and modelling explanations.

Agreement and disagreement both need covariance

Two analyses can share calibration stars, spectral templates, sky regions, sources, simulations or model assumptions. Their quoted uncertainties may consequently be correlated. Treating them as independent can overstate the precision of a joint result or misjudge the significance of their difference.

For a difference X − Y, the variance is Var(X) + Var(Y) − 2 Cov(X,Y). The covariance term may reduce or increase that difference uncertainty depending on its sign. The correct expression follows from the joint measurement model; choosing independence because the papers have different authors is not sufficient.

Likewise, agreement between methods is informative about the alternatives they test, but does not prove immunity to an assumption both share. The lensing example showed a transformation that preserved some image information; the background example showed a parallel component that preserved residual agreement. Combining genuinely complementary information can help address such specific ambiguities.

A discrepancy should motivate a traceable comparison: inspect definitions, units, calibrations, populations, selection, dependence and predictive assumptions. Additional measurements should be chosen for the differences they can resolve. More observations that respond identically to competing possibilities may improve precision without supplying identifiability.

Connect Astronomy with the separate Cosmology course

A usable handoff includes observations or their accessible provenance, response and processing descriptions, selection information, calibrated observables with joint uncertainty, and the rule used to compare them with predictions. It distinguishes withheld tests from development information and explains any limitations in coverage or source representativeness.

State the question the result answers. An observed redshift is not automatically a unique distance, a clustering amplitude is not a unique matter density, a lensing distortion is not a particle identification and a cleaned radiation map is not a cosmological parameter table. Each further interpretation needs its own conditional relationship.

The separate Cosmology learning map develops cosmic models, expansion histories, structure and parameter inference. This Astronomy chapter supplies observational links and their limits.

The next Astronomy chapter examines evidence for unseen mass and competing explanations, followed by other unresolved observational questions. Carry forward the same discipline: name what was measured, calculate what follows within a stated framework, and identify what additional evidence could discriminate between the remaining possibilities.

Worked examples

Calculate a conditional diagnostic

With z = 0.50, DA = 1000 Mpc and a proxy of 2500 Mpc, what is η in the stipulated transparent FLRW framework?

Predict DL = (1 + 0.50)² × 1000 = 2250 Mpc.
Divide the separate proxy by that prediction.
Simplify 2500/2250 to 10/9.

η = 10/9 ≈ 1.111111111; significance and cause are not determined by this exact arithmetic.

Check a supplied alternative

How can transmission q = 0.81 yield the same proxy while the true reference DL is 2250 Mpc?

Attenuation reduces flux by q while the proxy assumes transparent propagation.
The inferred distance scales as 1/√q.
Compute 2250/√0.81 = 2250/0.90.

The proxy becomes 2500 Mpc. This hypothetical alternative does not establish an observed attenuation mechanism.

Try the reasoning

Which separate luminosity-calibration ratio mimics a distance factor of 10/9 under transparent propagation?
Show the answer and reasoning
Lassumed/Ltrue = 100/81 ≈ 1.234567901.
Inferred luminosity distance scales with the square root of assumed luminosity.
Were attenuation and luminosity miscalibration both applied to obtain 2500 Mpc here?
Show the answer and reasoning
No.
Each is a separate alternative producing the same proxy ratio on its own.
Can results using the same calibration be assumed statistically independent because they were analysed separately?
Show the answer and reasoning
No.
Their covariance must follow the shared information and measurement model.
Is a fitted cosmic parameter a direct detector or map-pixel reading?
Show the answer and reasoning
No.
It summarizes a data comparison conditional on the chosen model and statistical assumptions.

What to carry forward

Preserve the transformation from detector records to conditional physical conclusions.
A consistency ratio can expose a disagreement without identifying its cause.
Combining evidence requires its calibration, selection, covariance and model dependencies, with separate course ownership retained.

References

Teaching resources and research supporting this lesson’s explanations and examples.