Paper detail

Extending the $Z^2_n$ and $H$ statistics to generic pulsed profiles

The search for astronomical pulsed signals within noisy data, in the radio band, is usually performed through an initial Fourier analysis to find "candidate" frequencies and then refined through the folding of the time series using trial frequencies close to the candidate. In order to establish the significance of the pulsed profiles found at these trial frequencies, pulsed profiles are evaluated with a chi-squared test, to establish how much they depart from a null hypothesis where the signal is consistent with a flat distribution of noisy measurements. In high-energy astronomy, the chi-squared statistic has widely been replaced by the $Z^2_n$ statistic and the H-test as they are more sensitive to extra information such as the harmonic content of the pulsed profile. The $Z^2_n$ statistic and H-test were originally developed for the use with "event data", composed of arrival times of single photons, leaving it unclear how these methods could be used in radio astronomy. In this paper, we present a version of the $Z^2_n$ statistic and H-test for pulse profiles with Gaussian uncertainties, appropriate for radio or even optical pulse profiles. We show how these statistical indicators provide better sensitivity to low-significance pulsar candidates with respect to the usual chi-squared method, and a straightforward way to discriminate between pulse profile shapes. Moreover, they provide an additional tool for Radio Frequency Interference (RFI) rejection.