Low voltage networks are used as guiding structures for Power Line Communication (PLC) for a long time. Models of such media are based on the multiconductor transmission line theory. In this paper, we apply a statistical approach in order to estimate the level of spurious signals on a low voltage distribution network. First, we describe how a test network can be modelled by means of the electromagnetic topology method used inside the CRIPTE code developed by the French Aerospace Lab (ONERA). We have chosen the classical TT earthing scheme. Then we explain how connected loads are characterized. After showing how uncertainty in the geometry of the cable bundle and variability of connected loads can have an effect on the spurious signals levels from 1 MHz to 100 MHz, the influence of both factors (geometry of cable bundle and variability of loads) has been studied. Then, the statistics of the extreme values of spurious signals has been derived. Last, the hierarchy of those factors has been discussed.