Skip to content

fit_nongrav="auto" cannot find a radial or out-of-plane non-gravitational acceleration #544

Description

@matthewholman

Two properties of the auto ladder mean it returns a gravity-only fit for objects with strong, published non-gravitational detections.

The ladder never tries A1 alone. It runs A2 -> A1A2 -> A1A2A3 (_AUTO_NONGRAV_LADDER), so a radial-dominated object has to reach the second rung, where A2 is already absorbing signal.

And the gate is on absolute fit quality, not on evidence for the parameter. A gravity-only fit at or below accept_reduced_chi2 = 1.5 is kept and no non-grav model is tried at all.

1I/'Oumuamua shows both. Its gravity-only fit gives reduced chi-square 0.637, so auto stops there — yet fitting A1 alone gives 4.4 sigma, within 1.7 sigma of the published 2.790e-7 +/- 3.57e-8 au/d^2. Fitting the full triplet is worse than useless: reduced chi-square is identical to three decimals across A1, A1A2 and A1A2A3 (0.588/0.589/0.589) while sigma(A1) inflates elevenfold and the sign flips.

The same shape of problem applies to the dark comets, where A3 is the significant term (Seligman et al. 2023, PSJ 4, 35) and A2 is not.

The ladder was designed for asteroids, where A2/Yarkovsky is expected and a failing gravity fit is the trigger, and it does that job. But an A1 rung, and a gate on parameter significance rather than reduced chi-square, would let it find the cases it currently cannot.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions