Statistical Power, the Bispectrum and the Search for Non-Gaussianity in the CMB Anisotropy

We use simulated maps of the cosmic microwave background anisotropy to quantify the ability of different statistical tests to discriminate between Gaussian and non-Gaussian models. Despite the central limit theorem on large angular scales, both the genus and extrema correlation are able to discriminate between Gaussian models and a semi-analytic texture model selected as a physically motivated non-Gaussian model. When run on the COBE 4-year CMB maps, both tests prefer the Gaussian model. Although the bispectrum has comparable statistical power when computed on the full sky, once a Galactic cut is imposed on the data the bispectrum loses the ability to discriminate between models. Off-diagonal elements of the bispectrum are comparable to the diagonal elements for the non-Gaussian texture model and must be included to obtain maximum statistical power.