In this paper an approach to analysis of dependence of Session Initiation Protocol server model with batch arrivals performance measures on batch size distribution is considered. Proposed approach employs non-parametric methods of statistical analysis. It is shown that there is statistical reliable dependence of performance measures, taken for signaling traffic analysis, on distance between distributions in definite norm. On the basis of proposed analysis elasticity coefficients were evaluated depending on distance between batch size distributions. This approach enables to get correction factors for estimation of these parameters in case distribution functions differ from uniform. © Springer International Publishing Switzerland 2016.