Eight of ten quadratic transit timing trends that pass single-transit removal fail a time group test under fixed errors in a 2022 catalog
We tested whether removal of time groups finds dependence that removal of one transit can miss. We used a fixed 2022 catalog with 4,096 distinct transit times for 106 systems. We assumed independent Gaussian errors with a minimum of 25.92 seconds. Groups have no gap greater than 90 days in the catalog ephemeris. A quadratic trend passes a removal test if every retained fit has linear-minus-quadratic BIC greater than 10 and keeps the full-fit sign. Of 21 nominal trends, ten pass removal of each single transit. Eight of these ten fail removal of at least one group. Six of the eight also have large coefficient shifts in a conservative Gaussian test with Holm correction across all 106 systems. Thus, loss of precision alone does not explain every failure under the stated model. WASP-12 keeps its known negative trend. TrES-3 passes both support screens but has a very poor quadratic fit. We provide all removals, selection records, source tables, numerical checks, sensitivity results, figures, and code. These are catalog diagnostics. They do not establish a new planet or a new case of orbital decay.
transit timing · exoplanets · period change · orbital decay · data quality · group removal · regression diagnostics · reproducible research
References
7- usesdoi:10.3847/1538-4365/ac545b
We use the public Tables 3 and 4 at a fixed repository commit. They supply the ephemerides, times, errors, and timing references.
- doi:10.3847/1538-4365/ad0847
This work uses single-transit removal to test period change candidates. It motivates our comparison with group removal.
- doi:10.3847/PSJ/ad3e80
This work shows how timing errors and a small set of observations can affect evidence for orbital decay.
- doi:10.1080/00401706.1977.10489493
This is prior work on regression influence diagnostics. We do not claim a new general deletion method.
- doi:10.3847/1538-3881/ac01dd
Table 5 prints the large TrES-3 time that we checked against the compiled catalog value.
- Free source timing offsets remove more than 90% of period change information in 43 of 115 systems in a 2022 transit catalog sci:sha256:361975de1be4f2ad9783a0385173d9d3213b82580ddc31835761a89ca6366c39
This related study tests source offsets in the same catalog. We compare a different source of model dependence.
- Optimized period change intervals are at least 10 percent shorter in 11 of 115 transit systems with 60-second source bounds and fixed Gaussian errors sci:sha256:51119d056b08148de736671a3a4a3641f1a82cdd85119182c07b128ae038105b
This related study tests bounded source offsets. Our new code tests changes after time groups are removed.
Cited by
5 publicationsPublic catalog entries on Sciscape.
- A quadratic fit to 2019–2024 TESS times does not resolve the reported period increase of Qatar-1 b
This study motivates tests of time group influence. We apply whole sector removal with independently written code and a different data set.
- supportsNine of twelve quadratic clocks selected from a 2022 catalog have larger later TESS errors than linear clocks under fixed timing errors
The external forecast results add evidence that apparent curvature in this catalog needs checks for poor fits and source faults. This study does not repeat its group removal test.
- usesTESS flux from 2021–2024 gives a TOI-2337 b period change of −0.29 ± 1.32 seconds per year with fixed noise estimates
Its time group tests motivate the sector removal checks used here.
- No constant period fits 50 published TOI-2015 b transit times within 139 minutes plus three times each quoted error
This work uses time-group removal to test timing fits. We apply the general idea to a different model, with calendar-year removal and deterministic interval limits.
- usesSix of eight selected early quadratic clocks have larger first TESS group mean errors than linear clocks in a 2022 catalog test with fixed errors
We build on its test of timing-trend stability in the same catalog. We add a separate early-literature to later-TESS prediction test. We do not reuse its code.