Fitted timing curvature has opposite signs in separate halves of 12 of 24 selected Kepler records

Curvature in part of a transit-time record may not continue in later data. We tested whether curvature in the first half of a Kepler record kept its sign and size in a separate second half. We used public Holczer catalog timings and a saved NASA object-status table. Each half needed at least 20 retained times over 500 days. Of 1,230 eligible records, 24 passed a first-half curvature test with Holm-adjusted p at most 0.01. Twelve had opposite fitted signs in the two halves. Thirteen rejected a common curvature coefficient at Holm-adjusted p at most 0.05. This test assumes independent Gaussian errors with one unknown scale. Five had opposite signs with curvature detected in both halves after adjustment over all half-record tests. For all 12 opposite-sign pairs, each fit keeps its sign when any one timing is removed. For seven pairs, each fit keeps its sign when any occupied 90-day block is removed. We provide each fit, each exclusion, fixed sensitivity runs, conditional controls, and independent numerical checks. The result is a diagnostic for this selected catalog sample. It is not a planet discovery, a population rate, or proof of physical orbital decay. Correlated timing signals and errors can invalidate the statistical model. The package includes exact numeric inputs, source hashes, code, results, figures, and offline reproduction instructions.

Kepler · transit timing · timing curvature · orbital period · disjoint time windows · public astronomical data · reproducible research