Constraining non-gaussianities in WMAP1 and WMAP2

We measure the 2-1 cumulant correlator power spectrum, a degenerate three-point statistic or integrated bispectrum, from the WMAP first and three year data releases (WMAP1 and WMAP2, respectively). We present a method of estimating these statistics using the pseudo-Cl-based SpICE estimator. We interpret the measurements in a maximum likelihood framework using theoretical predictions based on the simplest fNL model. All calculations are repeated on Monte Carlo simulations to obtain covariance matrices of our measurements. Application of the theory of random matrices revealed that the experimental covariances are consistent with a random matrix approximation. Finally, our χ2 analysis yields fNL = 22 ± 52(1σ) from WMAP2.