Pulsar glitch activity as a state-dependent Poisson process: parameter estimation and epoch prediction
Rotational glitches in some rotation-powered pulsars display power-law size and exponential waiting time distributions. These statistics are consistent with a state-dependent Poisson process, where the glitch rate is an increasing function of a global stress variable (e.g. crust-superfluid angular velocity lag), diverges at a threshold stress, increases smoothly while the star spins down, and decreases step-wise at each glitch. A minimal, seven-parameter, maximum likelihood model is calculated for PSR J1740-3015, PSR J0534+2200, and PSR J0631+1036, the three objects with the largest samples whose glitch activity is Poisson-like. The estimated parameters have theoretically reasonable values and contain useful information about the glitch microphysics. It is shown that the maximum likelihood, state-dependent Poisson model is a marginally (23-27 per cent) better post factum "predictor" of historical glitch epochs than a homogeneous Poisson process for PSR J1740-3015 and PSR J0631+1036 and a comparable predictor for PSR J0534+2200. Monte Carlo simulations imply that > 50 glitches are needed to test reliably whether one model outperforms the other. It is predicted that the next glitch will occur at Modified Julian Date (MJD) 57784 +/- 256.8, 60713 +/- 1935, and 57406 +/- 1444 for the above three objects respectively. The analysis does not apply to quasiperiodic glitchers like PSR J0537-6910 and PSR J0835-4510, which are not described accurately by the state-dependent Poisson model in its original form.