Audit an actual spectrum before treating a curve as an inference

1. Define the light and measurement

This lesson supplies an actual archived SDSS spectrum: public DR17 plate 358, MJD 51818, fiber 270, reduction RUN2D 26. The unchanged FITS source, complete 3819-row CSV, fixed selected windows, original flags and hash are downloadable. The source pipeline labels it QSO with redshift about 2.1095004 and ZWARNING 0; these are disclosed pipeline results, not a learner’s independent classification or proof of flawless data.

2. Connect physical and instrumental mechanisms

The COADD table stores LOGLAM, a base-ten logarithm of heliocentric vacuum wavelength in angstroms, and calibrated flux density in 10^−17 erg s^−1 cm^−2 Å^−1. Convert each actual row with λ=10^LOGLAM. The primary header’s COEFF0=3.5781 belongs to a wider grid, while this trimmed table begins near LOGLAM=3.5817000866. Applying that untrimmed offset to row zero would attach the wrong wavelength; source schema and actual row coordinates must agree.

IVAR is inverse variance in the source flux units; positive values permit formal uncertainty 1/√IVAR. AND_MASK retains bits present across all contributing exposures and OR_MASK bits present in any exposure. Different bits have different meanings, and a nonzero mask is not one universal severity category. The three displayed screens—positive IVAR alone, plus zero AND, or plus zero OR—are explicit pedagogical sensitivity comparisons, not a recommended blanket rejection rule for all SDSS science.

Four fixed windows contain 16 consecutive rows starting at indices 0, 1000, 2000 and 3000, selected without flux or quality filtering. A conditional inverse-variance mean is Σ(wF)/Σw, with formal error 1/√Σw only under independent-pixel and locally constant-flux assumptions. If no pixels survive, there is no estimate, not zero flux. Resampling covariance, calibration systematics and true spectral structure can invalidate that simple uncertainty or interpretation.

The overview is a fixed-stride source preview, not a rebinned or integrated spectrum, and cannot preserve every narrow feature. Source model and sky columns remain distinct from measured coadded flux. The learner audits provenance, calibration and quality rather than claiming a new astrophysical redshift fit. SDSS public-release data are public domain under its collaboration policy; source credit and links are retained, without implying endorsement.

Actual public-release spectrum: keep the source and reduction boundary

SDSS DR17 legacy plate 358, MJD 51818, fiber 270, RUN2D 26. Source pipeline class QSO and redshift 2.109500408 are disclosed labels, not this course’s independent fit. This cosmological redshift is not a peculiar velocity obtained by multiplying by c.

Displayed source window begins at zero-based COADD index 0. Screen 2: positive IVAR and OR_MASK zero. These are pedagogical sensitivity screens, not universal SDSS rejection recommendations. Original flags remain visible even when excluded.

Unmodified source fields; wavelength is derived as 10^LOGLAM, not the primary untrimmed COEFF0 grid

Index	LOGLAM	Vacuum Å	Flux	IVAR	AND	OR	Kept?
0	3.5817001	3816.8060	13.902831	0.091111913	0	0	yes
1	3.5818000	3817.6841	8.5142994	0.094205633	0	0	yes
2	3.5818999	3818.5623	4.5214238	0.10220698	0	0	yes
3	3.5820000	3819.4429	3.6619785	0.10624284	0	1048576	no
4	3.5820999	3820.3215	-1.6803676	0.11544725	0	1048576	no
5	3.5822001	3821.2025	2.7220814	0.11288712	0	0	yes
6	3.5822999	3822.0815	4.1783204	0.11130708	0	0	yes
7	3.5824001	3822.9629	11.099946	0.098191865	0	0	yes
8	3.5825000	3823.8424	6.0503359	0.10773776	0	0	yes
9	3.5826001	3824.7241	8.2220488	0.10753775	0	0	yes
10	3.5827000	3825.6040	15.479923	0.10107446	0	0	yes
11	3.5827999	3826.4841	12.612288	0.10355039	0	0	yes
12	3.5829000	3827.3665	10.919852	0.10480606	0	0	yes
13	3.5829999	3828.2469	11.277345	0.10511660	0	0	yes
14	3.5831001	3829.1297	2.8731112	0.11216995	0	0	yes
15	3.5832000	3830.0106	10.151416	0.098333664	0	0	yes

Original FITS · All 3819 source rows · All 64 fixed-window records · Source, units, selection, permissions and hash.

Original FITS SHA-256: 4d8d3f50a17b48b7aec7da068a298a48e955a0b8411ab206e63a85c1b63b2f3d

Source data credit: Sloan Digital Sky Survey, archived DR17. Public-release data policy treats these data as public domain; credit and links are retained. Original course explanations are not endorsed by SDSS. See quality documentation and individual flag meanings.

A formal mean error assumes independent pixels and constant local flux; covariance, systematics and real spectral structure are not supplied by this arithmetic. An empty retained set gives no estimate, not zero radiation.

3. Work the evidence

For the first fixed window, positive-IVAR and zero-AND screens retain 16 rows, weighted mean≈7.56334 and formal error≈0.773377 source flux units. Zero-OR retains 14 rows, mean≈8.58500 and error≈0.830387. The change is a policy sensitivity, not evidence that the object physically varied between these calculations.

4. Predict and investigate

If a stricter mask leaves no source pixels, should the displayed flux be zero or unavailable?

Verify the exact FITS identity/hash and selected row values. Reconstruct wavelength and one uncertainty from source columns, compare the declared screens, retain all original flags and explain which source/model assumptions prevent a new astrophysical inference.

Apply the evidence to a changed case

Compare the window beginning at index 2000 with zero-AND versus zero-OR selection. What does an empty estimate mean?

Retain the evidence frame, calculations and assumptions before opening public feedback.

Compare a worked response

Zero-AND retains 16 rows, giving mean≈3.14525 and formal error≈0.128657 source flux units. Zero-OR retains none, so mean/error are undefined. The original source rows still exist; the policy excludes them and does not establish zero received flux.

Source and boundaries

Primary source. Checked 30 October 2026. Original teaching and fictional numerical models are separate from the explicitly labeled actual SDSS archive; no direct observing, radiation experiment or instrument operation is requested.