Using a set of multifrequency cross-spectra computed from the three year WMAP sky maps , we fit for the unresolved point source contribution .
For a white noise power spectrum , we find a Q-band amplitude of A = 0.011 \pm 0.001 \mu K ^ { 2 } sr ( antenna temperature ) , significantly smaller than the value of 0.017 \pm 0.002 \mu K ^ { 2 } sr used to correct the spectra in the WMAP release .
Modifying the point source correction in this way largely resolves the discrepancy Eriksen et al .
( 4 ) found between the WMAP V- and W-band power spectra .
Correcting the co-added WMAP spectrum for both the low- \ell power excess due to a sub-optimal likelihood approximation—also reported by Eriksen et al .
( 4 ) —and the high- \ell power deficit due to over-subtracted point sources—presented in this letter—we find that the net effect in terms of cosmological parameters is a \sim 0.7 \sigma shift in n _ { \textrm { s } } to larger values : For the combination of WMAP , BOOMERanG and Acbar data , we find n _ { \textrm { s } } = 0.969 \pm 0.016 , lowering the significance of n _ { \textrm { s } } \neq 1 from \sim 2.7 \sigma to \sim 2.0 \sigma .