Removal of three nights decreases amplitude information by more than half for two of 57 HARPS candidates with fixed orbit curves under five white noise models
We tested how much a small set of observing nights can affect amplitude precision for 57 candidate signals and 251 signals matched to known planets in a published HARPS-RVBank search. We kept all published orbit curves fixed. We fitted signed amplitudes for all curves of each host and separate instrument offsets. We evaluated every one-, two-, and three-night removal with an exact matrix update. No single-night removal decreases information by half for any candidate under any tested noise model. In the main test, we added the source jitter variance to the quoted error variance. Removing three nights decreases conditional amplitude information by 52.8% for GJ613.0_1 at 170.51 days and by 56.7% for HD129814_2 at 82.02 days. These two cases also lose more than half under four alternative weight models. For GJ613.0_1, the same removal changes the fitted amplitude from 3.98 to 4.03 m/s. A circular version of its curve, with the same dates, period, mean longitude, and weights, loses at most 2.7% after any three-night removal. The test measures dependence on dates and adopted curves. It does not measure the chance that a planet is real. The archive contains the unchanged signed input publication, new code, complete result tables, figures, numerical checks, and instructions to repeat the work.
HARPS · radial velocity · planet candidates · observing cadence · eccentric orbits · conditional information · case deletion · reproducible analysis
References
6- usesA blind Keplerian search of 743 HARPS-RVBank v2 stars with 30 or more nights recovers 248 of 465 known planets and finds 57 signals (FAP < 0.1 %) with no known match that pass activity tests sci:sha256:17e2629023ac1bcbb507eb643afc93a73a34d8a9d1ba2357cc22bd7bc4f108cb
We use its final 57-candidate list, 251 known-planet matched signals, full precision orbit curves, fitted jitter, and nightly HARPS data. We preserve the complete signed source. We add an exact audit of every one-, two-, and three-night removal.
- usesdoi:10.1051/0004-6361/202348263
HARPS-RVBank v2 is the original public velocity data source. We use the nightly means preserved in the cited SCI search.
- doi:10.1051/0004-6361/201936686
HARPS-RVBank v1 and its SERVAL velocity methods underlie the source data.
- doi:10.1111/j.1365-2966.2004.08275.x
Prior work shows that eccentric orbit detection depends on sampling. Our test measures a conditional amplitude information loss for fixed published curves.
- doi:10.1093/mnras/stz1849
Prior work studies bias and robustness of eccentricity estimates under imperfect noise and signal models. We keep eccentricity fixed and do not test its statistical significance.
- doi:10.1080/00401706.1977.10489493
Observation deletion and influence in linear regression are established methods. Our reported variance ratio uses design geometry, rather than Cook distance.
Cited by
3 publicationsPublic catalog entries on Sciscape.
- usesFits to 57 fixed HARPS candidate curves give an 8.85 m/s DRS–SERVAL difference for GJ 579.2
This earlier audit used all 57 source curves to test amplitude information after night removals. We use the same source scope and test changes in the measured DRS minus SERVAL coefficients.
- 28 HIRES nights give similar fits for zero and the fixed 82-day HD 129814 curve with fitted long motion and Gaussian scatter
Its catalog abstract reports that removing three HARPS nights reduces the fixed-curve amplitude information by 56.7%. We use this as a prior warning and do not recompute it.
- 29 of 57 HARPS candidates reduce season error at all 12 settings with fixed periods and independent errors
This prior study tests night influence for the same source candidate set. It provides a distinct data-leverage check.