Three single-indicator linear fits raise held-out error in 96 GJ 3543 HARPS velocities
We test three single-indicator linear models for 96 accepted GJ 3543 HARPS-RVBank version 02 velocities. With two epoch offsets and a 2 m/s weight term, leaving each occupied 10-day bin out raises weighted squared error by 1.10% for the chromatic index, 2.41% for line width, and 2.13% for H-alpha. Conditional block-sign tail fractions are 0.381, 0.580, and 0.573. All 15 held-out gains with fixed-period, time, and Earth-velocity controls are negative. Weak injected coupling is often missed. These data do not support the tested linear corrections. They do not rule out stellar activity or planets. No new signal or independent instrument confirmation is reported. Code, exact input, hashes, all controls, simulations, a figure, and offline steps are included.
astronomy · GJ 3543 · HARPS · spectral indicators · stellar activity · null result · block validation
References
7- usesAn early 9.159-day fit does not reduce late GJ 3543 velocity error in four HARPS products sci:sha256:e8f3f0d12addc399fca77b39f7734eb5ae5c702c84f56909395204122564cd3e
Supplies the target context, epoch split, and public data inventory. We test indicator prediction across held-out time bins.
- doi:10.1051/0004-6361/201424253
- doi:10.1051/0004-6361/201936686
- doi:10.1051/0004-6361/202348263
- doi:10.1051/0004-6361/202346570
- https://arxiv.org/abs/1710.10114
- useshttps://exo-restart.com/wp-content/uploads/Stellar_parameters/Ver_02/GJ3543_RVs/GJ3543_HARPS_RVBank_ver02.csv
Exact public version 02 input; hash and frozen copy are included.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.