A new test using incidence data is developed for testing whether two or more groups have the same seasonal pattern. The method fits sine waves to the data with a fundamental period of one cycle per year, and has the possibility of using higher harmonics, when necessary, to adequately model the data. The seasonal pattern can, therefore, have an arbitrary shape. The method allows for different length time intervals and different size populations at risk in the time intervals. Maximum likelihood estimation, based on the Poisson distribution, is used to determine the parameters of the model. Likelihood ratio tests and Akaike's information criterion (AIC) are used to determine the number of harmonics, and to test hypotheses. This method has been used to test for seasonal patterns in the incidence of insulin-dependent diabetes mellitus (IDDM) in Colorado among persons aged 0-17 years. Comparisons of seasonal patterns are made between males and females, and three age groups, each controlling for the other effect as in analysis of variance. Other potential applications of this approach are also discussed. A basic program is available for an IBM-PC to carry out these analyses.