A non-parametric statistical model is constructed to directly relate The distribution of observed microlens timescales to that of the mass Function of the population from which the lenses are drawn, corrected For observational selection based on timescales and event amplifications. Explicit distributions are derived for microlensing impact parameters and maximum amplifications; both are shown to be statistically independent of all other parameters in the problem, including lens mass. The model is used to demonstrate that the narrow range of microlens timescales observed toward the Large Magellanic Cloud (LMC) is probably not consistent with lensing by a widely distributed spheroidal population of large velocity dispersion, as expected of a dark halo; but is consistent with lensing within a rotating thick disk. Poor numerical conditioning on the statistical connection between lens masses and event timescales, and small number statistics, severely limit the mass function information obtainable from current microlensing surveys toward the LMC.