1. Stacking gain
SNR_stack ≈ √16 × 15 = 4.00 × 15 = 60.0.
| Criterion | Points |
|---|---|
| Correct √N factor | 8 |
| Correct final SNR | 7 |
Worked Solutions
2. Cosmic-ray bias
Bias = 8000/16 = 500.0 counts added to the mean-combined pixel. A median or sigma-clipped combine would largely reject this single outlier instead.
| Criterion | Points |
|---|---|
| Correct bias value | 10 |
| Correct robustness statement | 5 |
3. Cadence adequacy
Samples across event = 3.4 h / 0.75 h = 4.5. About 5 samples is borderline: enough to detect the event but only marginally adequate to resolve ingress/egress shape in detail; a finer cadence would be preferable.
| Criterion | Points |
|---|---|
| Correct sample count | 8 |
| Reasonable adequacy judgment | 7 |
4. Nyquist-style check
Recovering a 5.2 h period needs a cadence well under half the period, i.e., well under 2.6 h. A 6 h cadence is coarser than that limit and risks aliasing. A 1 h cadence satisfies Δt ≪ T and would adequately sample the periodic signal.
| Criterion | Points |
|---|---|
| Correct Nyquist-style limit | 8 |
| Correct judgment and proposed fix | 7 |
Grading Notes
Award full marks for correct method, units, and a numeric result consistent with the arithmetic shown (allow reasonable rounding). Common partial-credit errors: omitting units, failing to combine uncertainties in quadrature, or overstating the confidence of a marginal (SNR 3–5) result.