An improved test for periodicity

Two widely used methods of testing for periodicity, phase dispersion minimization and epoch-folding are discussed. An analysis of variance approach demonstrates the close relationship between these two methods. It is also shown that the significance test sometimes used in phase dispersion minimization is statistically inaccurate and that the test used in epoch-folding is an approximation valid only for large sample sizes. A simple new test statistic is proposed, applicable to either epoch-folding or PDM, which is statistically sound for all samples sizes, and which is also more sensitive to periodicity than the test statistics previously used with these two methods.