One early glitch controls the 89 percent size-fit error gain on six later PSR J0631+1036 intervals

We test whether glitch size predicts the next waiting time in PSR J0631+1036. A size fit trained on ten early completed intervals gives 89.1 percent less mean squared error than an early constant fit on six later intervals in one Jodrell Bank catalog snapshot. One early pair has regression leverage 0.999. Removing it and refitting gives 27.9 percent more error than the new constant fit. Separate resampling of early and later pairs gives a gain range from -18.793 to 0.934. The six-interval test does not establish a reliable prediction law. Listed measurement errors have little effect under an assumed perturbation rule. Missing glitches, monitoring gaps and shared timing fits remain untested. We report no new glitch, source or planet. The package contains original code, aggregate results and a figure. External catalog records are excluded. Exact paths, hashes and retrieval instructions are included.

astronomy · pulsar glitches · PSR J0631+1036 · prediction · regression leverage · null result