Two noise mixtures give a 0.29 to 1.54 nat gain for the 50.77-day term in 151 older GJ 887 velocities
We test whether the reported 50.77-day GJ 887 signal predicts 151 older velocities from HIRES, PFS, and UCLES. We fit public HARPS and ESPRESSO data with linear planet terms, proper Gaussian coefficient priors, and fixed reference activity parameters. We remove instrument offsets with orthonormal contrasts and score the full test vector. We average over 12 specified legacy noise choices and a local period band from 50.52 to 51.02 days. The joint prediction gain is 1.543 nats with separate activity processes and 0.287 nat with shared activity. HIRES and PFS give positive gains, while UCLES gives a negative gain. The known 9-day and 22-day signals give joint gains of 8.81 to 17.10 nats in control tests. A longer activity correlation time reverses the shared-activity result in a reference-noise check. Exact Gaussian predictions and 10,000 draws per model show limited expected information for the 50-day term at the observed test times. This is a retrospective instrument test. The older data helped select the 2020 candidate, and later work supplies the periods and noise reference. The result gives weak support under the two stated mixtures. It does not confirm a planet. The package preserves all five source tables, a complete row ledger, code, figures, and numerical checks.
GJ 887 · radial velocity · exoplanets · predictive density · Gaussian process · instrument validation · stellar activity · public astronomical data
References
8- usesdoi:10.1051/0004-6361/202554984
This article supplies the reported planet periods and the four-planet activity parameter reference. It omits the three legacy instruments from its fit.
- usesdoi:10.26093/cds/vizier/bc-p3/ljd
The CDS tables supply all HARPS and ESPRESSO source observations used here. We preserve the released tables and state the sampling difference from the article.
- usesdoi:10.6084/m9.figshare.12271568.v2
This version of the public archive supplies the 75 HIRES, 38 PFS, and 38 UCLES night means. We preserve these source tables and their archive paths.
- doi:10.1126/science.aaz0795
This article reported planets b and c and the uncertain 50-day signal. Its data history makes the present test retrospective.
- doi:10.3847/1538-3881/ae0b62
This earlier analysis used HARPS and HIRES observations of GJ 887 and did not recover the 50-day term in its model.
- usesdoi:10.1093/mnrasl/sly227
This article describes the HIRES hardware change and nightly zero-point correction. It supports the stated treatment of the released HIRES data.
- usesA period average reduces the GJ 887 candidate forecast gain under fixed noise assumptions sci:sha256:f44c4414fade1d53edad73d53323c3456a9934e8d657df95e67d1e2ddd7b7c43
We use its public-table sampling rule and its treatment of local period uncertainty. Its tested GJ 887 candidate has a different period, 2.21661 days.
- usesThe fixed 53.647-day GJ 251 term improves joint infrared prediction under three tested noise models sci:sha256:510340eb9c55963e67328665112620aa034b1189ca684f03008433dd660ff599
We use its method of instrument prediction with orthonormal offset contrasts, proper Gaussian signal priors, and separate or shared activity. Our implementation and GJ 887 data are supplied here.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- A 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
Our blind search of GJ 887 finds the known 9.26-day and 21.79-day planets. It does not detect the 50.77-day signal (S = 18.1). Our 39.21-day signal agrees with the published rotation signal.
- usesFixed correlated HARPS noise moves 60-75-day GJ 581 forecast scores toward zero in 54 later velocities
This GJ 887 study informs joint instrument forecasts and offset removal with orthonormal contrasts. It also informs planet controls and conditional simulations.