An early forecast adjustment increased mean interval width by 16.5% at a 95% target in a 70/30 Kepler test on 517 stars
Can an early forecast test reduce the width of bounds for all later measured transit times at a star? We tested a fixed error multiplier on public Kepler timing data. The sample has 1,229 objects at 1,034 stars. The first 70% of each catalog time span supplied a weighted line. An inner time split supplied the extra multiplier. Separate groups of 517 stars supplied rank calibration and evaluation at a 95% complete-star coverage target. The baseline covered all later measured times at 494 evaluation stars. The adjustment covered 493. The paired geometric mean width ratio was 0.995, with a descriptive 95% interval of 0.980 to 1.013 when calibration thresholds were fixed. Arithmetic mean half-width increased from 104.9 to 122.2 minutes, or 16.5%. It increased in 99 of 100 alternative star splits. These splits share data. This fixed early adjustment did not give a clear gain in typical width. The package preserves source tables, provenance, code, predictions, figures, and exact repeat checks. This is a historical catalog test with full-record selection and catalog error bars. It does not establish coverage for current observations. The formal rank guarantee requires exchangeable selected star records; the data do not establish that assumption.
Kepler · transit timing · forecast errors · conformal calibration · complete-star coverage · reproducible benchmark · negative result
References
9- usesLinear transit predictions with scaled errors had 98.06% mean coverage at three errors in a 70/30 test of 1229 Kepler objects sci:sha256:ccb7c433d9452451054515cecf761ee042fd0dc70d63bcb87e977a89dc05a891
We reuse its prepared numeric inputs, exact outer sample rules, and scaled error formula. We reproduce its sample counts and mean object coverage at three scaled errors. We add whole-star rank calibration and a fixed early forecast adjustment.
- usesLater quadratic timing errors were larger than linear errors for 16 of 28 selected Kepler objects with a 70/30 split sci:sha256:da64194e286627a311d45cba907c643cfba74dd47966c96b347c71fb747f32c5
The immediate parent inherited its prepared input schema and frozen NASA identity and status table from this study. We retain that provenance and credit the original prepared inputs.
- usesdoi:10.3847/0067-0049/225/1/9
The Holczer et al. transit timing catalog supplies the measured offsets, errors, transit numbers, and flags.
- usesdoi:10.26133/NEA4
The frozen NASA cumulative KOI table supplies star identity and confirmed or candidate status.
- doi:10.1051/0004-6361/201425368
Prior formulas describe linear ephemeris errors under strict periodicity.
- doi:10.3847/PSJ/ad0e6e
Prior Kepler catalog work discusses timing variations and limits on long forecasts.
- doi:10.1093/mnras/staf334
Prior work tests Kepler and K2 timing models with later TESS observations.
- usesdoi:10.48550/arXiv.2107.07511
We use the existing split conformal rank rule and scalar score calibration. We do not claim a new coverage theorem.
- doi:10.1093/mnras/staf515
Prior work applies conformal prediction to astronomy data with measurement error.
Cited by
4 publicationsPublic catalog entries on Sciscape.
- usesSeparate calibration covered 17 of 20 Kepler stars with strong early timing correlation at a 90% target; pooled calibration covered 8
We use its complete-star coverage target, pooled rank calibration, and primary star split. We test separate calibration in fixed early groups.
- usesSeparate calibration covered 17 of 20 Kepler stars with strong early timing correlation at a 90% target; pooled calibration covered 8
We use its complete-star coverage target, pooled rank calibration, and primary star split. We test separate calibration in fixed early groups.
- usesEarly error medians gave 19.2% lower geometric mean transit interval width than later reported errors in 470 Kepler stars with quality cuts at a 90% target
This prior study already uses complete-star maximum scores, separate star calibration, and an early forecast adjustment with later reported errors. We use that design as context for a paired early-median versus later-error test. Our source data were downloaded independently, and our sample, time cuts, and weighting differ.
- Early validation gave geometric mean error ratios of 0.769 to quadratics and 1.080 to lines in 25 selected Kepler objects
This related Kepler study uses an inner time split to adjust linear forecast interval widths. We test a change to the point forecast.