We present a method for deriving a smoothed estimate of the peculiar velocity field of a set of galaxies with measured circular velocities $\eta\equiv { \rm log} \Delta v$ and apparent magnitudes $m$. The method is based on minimizing the scatter of a linear inverse Tully-Fisher relation $\eta= \eta(M)$ where the absolute magnitude of each galaxy is inferred from its redshift $z$, corrected by a peculiar velocity field, $M \propto m - 5\log(z-u)$. We describe the radial peculiar velocity field $u({ \bf z})$ in terms of a set of orthogonal functions which can be derived from any convenient basis set; as an example we take them to be linear combinations of low order spherical harmonic and spherical Bessel functions. The model parameters are then found by maximizing the likelihood function for measuring a set of observed $\eta$. The predicted peculiar velocities are free of Malmquist bias in the absence of multi-streaming, provided no selection criteria are imposed on the measurement of circular velocities. This procedure can be considered as a generalized smoothing algorithm of the peculiar velocity field, and is particularly useful for comparison to the smoothed gravity field derived from full-sky galaxy redshift catalogs such as the IRAS surveys. We demonstrate the technique using a catalog of ``galaxies" derived from an N-body simulation. Increasing the resolution of the velocity smoothing beyond a certain level degrades the correlation of fitted velocities against the velocities calculated from linear theory methods, which have finite resolution,
Categories: astro-ph