A statistical approach for the sedimentation of prognostic precipitating species inside a microphysical scheme is proposed. The unique fall speed of precipitation used by a typical advective scheme is replaced by three probabilities of transfer associated with the following three types of precipitation: (i) precipitation present in the layer at the beginning of the time step; (ii) precipitation coming from the layer above which crosses the layer under consideration and (iii) precipitation locally produced (or destroyed) during the time step. The use of a decreasing exponential Probability Distribution Function (PDF) as basis for these probabilities ensures that the normalized probabilities do not depend on the origin of the precipitation. The main advantage of this characteristic is that the sedimentation process can now be computed in a single vertical loop. Other functions can also be used as basis for the probabilities and the case with a step function is discussed. The performance of the proposed PDF-based sedimentation scheme is validated in an academic single column model and in a full 3-D framework.