Title:
Inferring the velocity distribution of nearby stars from Hipparcos data

Abstract:
We present a three-dimensional reconstruction of the velocity
distribution of nearby stars ($\lesssim 100$ pc) using a maximum
likelihood density estimation technique applied to the
two-dimensional tangential velocities of the stars. The underlying
distribution is modeled as a mixture of Gaussian components.  The
algorithm reconstructs the error-deconvolved distribution function,
even when the individual stars have unique error and missing-data
properties. We apply this technique to the tangential velocity
measurements from a kinematically unbiased sample of 11,865 main
sequence stars observed by the Hipparcos satellite. We explore
various methods for validating the complexity of the resulting
velocity distribution function, including criteria based on Bayesian
model selection and minimum coding inference, as well as how
accurately our
reconstruction predicts the radial velocities of a sample of stars
from the Geneva-Copenhagen survey. Thus, we can quantify the
information content of the radial velocity measurements, which is
interesting in the light of the upcoming Gaia mission. We find that
the mean amount of new
information gained from a radial velocity measurement of a single
star is significant, which strongly argues for a complementary radial
velocity survey to Gaia. We also confirm the existence of  ``moving
groups'' in the velocity
distribution of the disk of the Galaxy, quantifying their statistical
significance for the first time. We find that the color-magnitude
diagrams of most of the moving groups are inconsistent with being
trails of evaporating, young clusters, which favors their
interpretation as being due to dynamical resonances or
non-axisymmetry and time-dependence of the disk potential.