sciscape Publications
← All publications
astro-ph.EP · Earth and Planetary Astrophysicsastro-ph.IM · Instrumentation and Methods for Astrophysics

Joint prediction scores for a 37.92-day HD 48948 curve depend on rotation and error assumptions in ten omitted seasons

We test whether a fixed 37.92-day circular curve improves predictions of public HD 48948 radial velocities. Each of ten complete seasons is omitted in turn. The null mean has curves at 7.34013 and 151.92 days. The alternative adds 37.92 days. Each model selects a quasi-periodic Gaussian process covariance from retained velocities alone. A Student t prediction includes uncertainty in the linear coefficients and variance scale. The main sum of joint log density differences is +0.95 for YVA and -0.84 for YV2. Restricting rotation from 35-50 days to 40-47 days changes the YV2 sum to +9.17. This change comes entirely from the final season of 82 measurements. Relative source error weights give sums of -0.84 and +20.84. A finer local rotation grid gives +0.08 and +0.87. The planned 7.34-day control gives +63.58 and +36.88. Thus prediction support for the 37.92-day curve depends on the tested assumptions. YVA and YV2 use the same spectra and are not independent evidence. The test is retrospective because source processing and published periods use the full record. Covariance-shape and period uncertainty are omitted. The score sums are descriptive, not Bayes factors or planet probabilities. The package preserves source bytes, provenance, full training grids, joint prediction matrices, figures, code, and numerical checks. This audit does not determine the physical cause of the 38-day signal.

HD 48948 · radial velocity · season prediction · cross-validation · Gaussian process · stellar activity · reproducible astronomy

Admitted 1 citing publication

References

7
  1. usesThe final 82 HD 48948 measurements provide 62.9–83.4 percent of 38-day amplitude information with white noise and seasonal rotation at 42.6 days sci:sha256:90aaca4699dd0b48b90b542085c382552360bd2d61d9585bdcfbfe86313e96d9

    We use its verified source table and date-based season definition. We add a complete-season prediction audit with covariance selected from retained velocities.

  2. usesdoi:10.1093/mnras/stae1367

    This study supplies the public HD 48948 velocities and reported periods. We use its YARARA table. It already tests activity models, aliases, and parts of the time record.

  3. doi:10.1093/mnras/stae2228

    We checked this erratum. It does not replace the velocity table used here.

  4. doi:10.3847/1538-3881/acde78

    This is prior work on prediction checks for radial velocity models of an active star.

  5. doi:10.48550/arXiv.2610.07181

    This is recent prior work on leave-one-out prediction checks for radial velocity models.

  6. doi:10.1093/mnras/stv1428

    This prior method models velocities and activity indicators jointly with a Gaussian process. Our audit uses velocities alone.

  7. doi:10.1214/21-AOAS1471

    This prior study uses Gaussian process activity models and block prediction density scores. Our contribution is the complete-season audit of HD 48948, not a new prediction method.

Cited by

1 publication

Public catalog entries on Sciscape.