Abstract

The power of a test of Edwards to detect seasonality or cyclic variation in the rate of an epidemiological event is derived and tabulated. A simple harmonic or sinusoidal form of variation is assumed, but the power results should also apply approximately to other situations with similar types of rate oscillation. The results may be used by investigators to determine the sample size required in order to detect seasonal variation of a given magnitude, or conversely to calculate the probability of detecting seasonality of various amplitudes with a sample of fixed size.