# Astronomical inference, selection and multi-messenger data

> A detected sample is shaped by observing limits and selection criteria. It need not represent the full source population without correction.

Course: [Astrophysics, cosmology and inference](https://studybrix.com/en/courses/astrophysics-and-cosmology/index.md)
Level: advanced
Prerequisites: Probability, data analysis and uncertainty.

## Notes

### Selection

A detected sample is shaped by observing limits and selection criteria. It need not represent the full source population without correction.

### Messengers

Electromagnetic radiation, gravitational waves and particles can reveal complementary processes. Association requires consistent timing, direction and false-coincidence analysis.

### Model comparison

A good fit is not proof that a model is uniquely correct. Compare alternative explanations, residuals and out-of-sample predictions when possible.

### Reproducibility

Record data versions, calibration, selection rules and analysis code. Report uncertainty and distinguish measured quantities from derived estimates.

## Practice questions

### 1. Why can a flux-limited survey overrepresent luminous sources?

A. They can be detected over larger volumes
B. All sources have the same luminosity
C. Distances are irrelevant
D. Faint objects cannot exist

Answer: They can be detected over larger volumes

Explanation: Detectability depends on luminosity and distance.

### 2. A survey detects 30 objects from 100 known eligible injected test signals. Find measured recovery fraction.

Answer: 0.3

Explanation: Recovery fraction is 30/100 for this injection sample.

### 3. What supports associating two messenger events?

A. Matching file names only
B. A statement without uncertainty
C. Identical website colours
D. Consistent timing and location with coincidence analysis

Answer: Consistent timing and location with coincidence analysis

Explanation: Association requires evidence against accidental coincidence.

### 4. Why is a nondetection not always evidence that a source does not exist?

Sample answer / marking guidance: The signal may lie below sensitivity or outside coverage. Upper limits and selection effects describe what the observation actually constrains.

Explanation: Credit detection thresholds and observational coverage.

### 5. A model predicts 40 counts and 44 are observed. Find the signed residual observed minus predicted.

Answer: 4 counts

Explanation: 44−40=4; significance needs an uncertainty model.

### 6. Explain why the size of a residual alone does not establish statistical significance.

Sample answer / marking guidance: Its importance depends on noise, uncertainty, correlations and the number of comparisons. A residual needs a reference distribution or appropriate uncertainty scale.

Explanation: Credit uncertainty and context.

### 7. Which item belongs in a reproducible astronomical analysis?

A. Only a claim of certainty
B. Only the final rounded number
C. Data version and selection rules
D. Only an unlabeled screenshot

Answer: Data version and selection rules

Explanation: Inputs and decisions are necessary to reproduce the inference.

### 8. Design a brief check for a claimed population trend in an observed catalogue.

Sample answer / marking guidance: Inspect completeness and selection, compare uncertainty-aware models, check calibration and confounders, and test whether the trend survives reasonable alternative selections.

Explanation: Credit a concrete selection check and a robustness check.

## Further reading

- https://science.nasa.gov/universe/

Original AI-assisted educational draft; not independently teacher-reviewed. References support further study, not copied exercises, affiliation or curriculum accreditation.