On the validity of Davies' test for periodicity

Abstract

Davies' test (1990, MNRAS, 244) for periodicity, which is based on the phase dispersion minimization method, is shown to be of limited application. In his derivation of the null distribution of the test statistic (i.e. the distribution under the null hypothesis of no periodicity), Davies assumes that the period to be tested is known a priori. Consequently, although the test can be used to test periods, that are identified by conducting a period search of the data. In the latter case, Davies uses adjustment based on Bonferroni's Inequality to compensate for the fact that multiple trial periods have been examined. Until a better solution to this difficult problem has been found, the use of the non-parametric randomization test described by Nemec & Nemec (1985, Astron. J., 90) is recommended