Abstract
A one-parameter family of tests for periodicity in a time series is proposed that contains Fisher's test as a special case. Some tests in this family have substantially larger power than Fisher's test against many alternatives, yet retain comparable power against those alternatives for which Fisher's test is optimal. A Monte Carlo power study establishes this and also provides a basis for the selection of a test from this family. Critical values are obtained by using a duality with the problem of covering a circle with random arcs.