Using a discrete wavelet based space-scale decomposition (SSD), the spectrum of the skewness and kurtosis is developed to describe the non-Gaussian signatures in cosmologically interesting samples. Because the basis of the discrete wavelet is compactly supported, the one-point distribution of the father function coefficients (FFCs) taken from one realization is a good estimate of the probability distribution function of the density if the ``fair sample hypothesis holds. These FFC one-point distributions can also avoid the constraints of the central limit theorem on the detection of non-Gaussianity. Thus the FFC one-point distributions are effective in detecting non-Gaussian behavior in samples such as non-Gaussian clumps embedded in a Gaussain background, regardless of the number or density of the clumps. We demonstrate that the non-Gaussianity can reveal not only the magnitudes but also the scales of non-Gaussianity. Also calculated are the FFC one-point distributions, skewness and kurtosis spectra for real data and linearly simulated samples of QSO Ly$alpha$ forests. When considering only second and lower order of statistics, such as the number density and two-point correlation functions, the simulated data show the same features as the real data. However, the the kurtosis spectra of samples given by different models are found to be different. On the other hand, the spectra of skewness and kurtosis for independent observational data sets are found to be the same. Moreover, the real data are significantly different from the non-Gaussianity spectrum of various posssible random samples. Therefore the non-Gaussain spectrum is necessary and valuable for model discrimination.
Download