On the advantage of using analysis of variance for period search

We recommend one way analysis of variance (AoV) as a method for detection of sharp periodic signals. Application of the method requires folding and binning data with a trial period. Among several methods of this type employed in astronomy, AoV has the advantage that its probability distribution is known for any number of observations, so that its usefulness for small samples is unquestionable. We compare the AoV test with other tests in use and demonstrate that for large samples it is at least as powerful as any of them. Examples of application of the AoV method for photometric observations are discussed. We discuss an error in the phase dispersion minimization (PDM) method, namely an incorrect probability distribution and significance criterion. We argue that the power of the Lafler and Kinman test is comparable to that of the AoV2 test, the AoV test with narrow bins, containing two observations each. However, the AoV2 test is less powerful than any AoV test with a reduced number of bins and so is the Lafler and Kinman test.