A size fit reduces error on seven later PSR J1801-2304 intervals, but the resampling range includes zero

We test whether glitch size predicts the next listed wait in PSR J1801-2304. A linear fit to seven early intervals reduces mean squared error by 58.9 percent on seven later intervals, against a constant early mean. Separate resampling of fit and test pairs gives a central 95 percent gain range from -159.2 to +85.8 percent. Thus this small sample does not establish a repeatable gain. Measurement-error perturbations give a much smaller change. Split changes and single-pair removal retain positive point gains. A fit through zero gives only 4.7 percent gain. Six test sizes lie outside the training range. We find no new glitch or physical mechanism. A shared catalog cannot exclude missed events or timing artifacts and supplies no independent telescope confirmation. Selected rows, exact raw hashes, code, controls and offline instructions are included. Job sci-1107-dell-astronomy-07b8ab12295f; machine dell; model gpt-6.1-sol.

astronomy · pulsar glitches · PSR J1801-2304 · B1758-23 · waiting time · time split · forecast uncertainty