Selection Effects in Periodic X-Ray Data from Maximizing Detection Statistics

The Neutron Star Interior Composition Explorer (NICER) records data of exceptional quality on the energy-dependent X-ray pulse profile pulsars. However, in searching for evidence pulsations, Guillot et al. (2019) introduce a procedure to select an ordered subset that maximizes detection statistic (the H-test). I show this can degrade subsequent analyses using idealized model with stationary expected count rates from both noise and signal. Specifically, data-selection biases inferred mean rate be too low, pulsation amplitude high, size these scales strongly amount is rejected true signal amplitude. also alters null-distribution H-test rendering nominal significance estimates overly optimistic. While does not capture all complexities real NICER data, it suggests could important NICER's observations J0740+6620 other faint pulsars (observations J0030+0451 are likely less affected). estimate selection effects may bias $\mathcal{O}(10\%)$ average modulation depth lightcurves like J0740+6620's, as large $\mathcal{O}(50\%)$ fainter change single dataset shift between $-5\%$ $+20\%$. This imply lower limit J0740+6620's radius slightly larger than should be, although preliminary investigations suggest constraints $\mathcal{O}(1\%)$ data.