An original ElectroMagnetic simulation methodology based on the hybridization of deterministic and statistical models is proposed to study wireless signal propagation inside a train. The coupling between the EM methods is performed with the Kron's networks formalism to significantly optimize the compromise between simulations accuracy and computation resources. The proposed hybrid model offers promising applications concerning parametric analysis of wireless interferences in complex systems such as train. For example, one can quickly identify EM impacts of sensitive parameters such as antennas, windows, seats and passengers in a reverberant environment (a carriage) in order to optimize the communication architecture during pre-design phases of train. Finally, modeling results are compared with preliminary experiments.