Optimal estimator for the amplitude of the bispectrum from infrared clustered sources

We devise a fast and optimal estimator for the amplitude of the bispectrum of the clustered point sources constituting the cosmic infrared background (CIB). We show how this estimator can account for the cases of partial sky coverage and inhomogeneous noise. The expected detection significance is presented in terms of signal-to-noise, showing that the CIB bispectrum will be undetectable below 220 GHz with a Planck-like experiment; in contrast, detection may be achieved at or above 220 GHz, especially if the CMB is removed. We also show how this estimator can be combined with estimators of unclustered sources and CMB non-Gaussianity to build up joint robust constraints. On the one hand, we find that, for a Planck-like experiment, CMB non-Gaussianity estimation can be decoupled from point source contributions, if the brightest sources are masked. On the other hand, we find that the estimation of clustered and unclustered sources non-Gaussianity are strongly coupled, which diminishes their separate detection significance.