Professor Pedro Ferreira

Abstract:

To correctly analyse data sets from current microwave detection technology, one is forced to estimate the sky signal and experimental noise simultaneously. Given a time-ordered data set we propose a formalism and method for estimating the signal and associated errors without prior knowledge of the noise power spectrum. We derive the method using a Bayesian formalism and relate it to the standard methods; in particular we show how this leads to a change in the estimate of the noise covariance matrix of the sky signal. We study the convergence and accuracy of the method on two mock observational strategies and discuss its application to a currently-favoured calibration procedure.

Abstract:

I review our recent claim that there is evidence of non-Gaussianity in the 4 Year COBE DMR data. I describe the statistic we apply, the result we obtain and make a detailed list of the systematics we have analysed. I finish with a qualitative understanding of what it might be and its implications.