Testing that a stationary time series is Gaussian (Q1102680)

The paper introduces procedures for testing the composite hypotheses that stationary stochastic processes are Gaussian. A recent simulation study indicates that the usual Pearson chi-square test can be greatly excessive when applied to correlated data. For time series with the same assumptions on the rate of convergence to zero of its covariance functions, the author introduces a test based on an empirical characteristic function. In contrast with the bispectral density test this one relies on quadratic forms in deviations of certain sample statistics from their population counterparts, minimized with respect to the unknown parameters. Canadian lynx and Wolfer sunspot data are used to illustrate the applications of the introduced test.