A fixed SHO filter leaves 3.66% to 5.22% of added 35.68-day signal amplitudes in 172 GJ 832 HARPS velocities
We tested how a fixed stellar activity filter changes circular orbit signals in 172 public HARPS velocities for GJ 832. We fixed a stochastic harmonic oscillator covariance at published parameter values. We fit two instrument offsets and a circular outer orbit. At 35.68 days, the residual sine fit retains 3.66% to 5.22% of an added signal amplitude over all phases. With the correct residual noise covariance, this test has the same information as a joint fit. Adding a signal after the filter changes the experiment. At a 1% false alarm rate for one known period, 90% recovery requires 1.449 to 1.506 m/s before filtering and 0.0543 to 0.0768 m/s after filtering. The ratio at the same phase is 19.2 to 27.3. Direct additions and 30,000 Gaussian trials verify the calculation. We preserve code, source data, provenance, phase bounds, matrices, simulation scores, and figures. These are conditional calibration results. The fixed noise model does not fully describe the observed velocities. The results do not establish a planet or a planet exclusion limit.
GJ 832 · HARPS · radial velocity · stellar activity · Gaussian process · signal injection · detection power · fixed covariance · SHO kernel · reproducible research
References
6- usesdoi:10.1051/0004-6361/202243063
We use the public CDS velocity table J/A+A/664/A64/rv.dat and the published outer period. We keep all 172 HARPS measurements.
- usesdoi:10.1051/0004-6361/202347902
We use the median GJ 832 SHO and white noise parameter values in Table 5 for a fixed covariance experiment. Section 5.3 motivates the comparison of signal additions. We do not reproduce its data extraction or planet limits.
- doi:10.1088/0004-637X/791/2/114
The earlier 35.68-day planet claim sets the primary test period. We do not select this period from the present fit.
- doi:10.3847/1538-3881/aa9332
This paper gives the celerite SHO covariance method. We implement the kernel directly and verify it with an independent state-space calculation.
- doi:10.1093/mnras/stw2261
Earlier work discusses signal changes in separate activity fits and joint fitting. We do not claim that the linear projection algebra is new.
- A six-index linear activity fit reduces 42.38-day signals at all phases in the published HD 26965 NEID data sci:sha256:9270d06a9645dd96e612802402610e8a5cc3419530c0d62ec76525f44a17c4b6
This prior SCI study gives exact phase bounds for fixed linear activity fits to HD 26965. We extend the comparison to a fixed GP filter and paired detection powers. We do not reuse its data or code.
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 832 (180 nights) finds only GJ 832 b and no 35.68-day signal at FAP < 0.1 %.
- A fixed activity covariance gives a 1.40 ± 0.44 m/s amplitude for the 50.81-day GJ 674 template in 248 HARPS nights
This GJ832 filter study motivates a direct amplitude audit of an observed GJ674 candidate. We use none of its code or data. The new contribution is the GJ674_3 fit, its known-orbit controls and its covariance sensitivity set.