Lecture 10: Survey Selection and Malmquist Bias

ASTR 476: Observational Cosmology · 2026-09-27

Driving question: Which objects never enter the catalog?

Learning objectives

  1. Define a selection function and distinguish it from an observed distribution.
  2. Explain Malmquist and Eddington biases with the direction of the effect stated.
  3. Design a completeness diagnostic that uses injection/recovery or external calibration.
Terms · 8 min

Terms for This Lecture

Selection function

Definition: The probability that an object with specified latent and observed properties enters the analyzed sample.

Why it matters here: It is required to answer Which objects never enter the catalog? and to distinguish direct observation from cosmological inference.

Relationships: Read it with Malmquist bias, the model \(p(detected|x) = S(x) p(x) / integral S(x')p(x') dx'\), and the evidence statement The release documentation and published sample counts are cited; no survey completeness map or detection efficiency curve is bundled.

Boundary or common confusion: An observed histogram cannot by itself identify completeness because it mixes cosmological volume, population evolution, and selection.

Slide 1 · 8 min

Catalogs are filtered populations

Draw the path from astrophysical population to detection, quality cuts, and analyzed sample.

Slide 2 · 10 min

Flux limit changes the mix

Explain why intrinsically brighter objects dominate at larger distance and how that can bias inferred luminosities.

Slide 3 · 12 min

Measurement scatter at a threshold

Show Eddington bias when more objects scatter upward from a steep population than downward.

Slide 4 · 12 min

Redshift and cadence selection

Discuss follow-up probability and the coupling between transient light curves and observing schedule.

Slide 5 · 10 min

Read the excerpt carefully

Show the actual stratified redshifts and contrast them with an unknown survey selection curve.

Slide 6 · 8 min

Correct, then propagate

Explain forward modeling of selection and why truncating a likelihood after fitting is generally wrong.

Survey Selection and Malmquist BiasThe release documentation and published sample counts are cited; no survey completeness map or detection efficiency curve is bundled.redshift zHDcount in selected excerpt0.000.250.500.751.001.251.501.752.00fixed-target redshift selectionObserved Pantheon+ excerpt redshifts; histogram is shaped by our teaching selection rule
The release documentation and published sample counts are cited; no survey completeness map or detection efficiency curve is bundled.

Quantitative anchor

The 18 Pantheon+ redshifts were selected near fixed targets, so their histogram cannot estimate the parent survey completeness.

\[p(detected|x) = S(x) p(x) / integral S(x')p(x') dx'\]

Interpretation and limits

A flux-limited survey samples a population through a selection function. Detection thresholds, cadence, color cuts, spectroscopic follow-up, and analysis cuts shape the observed redshift and luminosity distributions.

Boundary: An observed histogram cannot by itself identify completeness because it mixes cosmological volume, population evolution, and selection.

Evidence: The release documentation and published sample counts are cited; no survey completeness map or detection efficiency curve is bundled.

Active check: Explain forward modeling of selection and why truncating a likelihood after fitting is generally wrong.

ASTR 476 · Lecture 10 · 60-minute module · Survey Selection and Malmquist Bias