Computer simulation occupies an increasingly important place between theory and experiment. Simulation on large computers (Monte Carlo sampling) makes it possible to realize the so-called synthetic analysis of complex phenomena in which the interactions between fundamental constituents are known but not the global behaviour of the system. This paper introduces a global protocol for the comparison of model simulations with experimental results. Basic statistical tools as well as a recursive inversion procedure and multivariate analysis techniques are described. This protocol is relevant in all fields of physics dealing with event generators and multiparametric experiments. In the present work, the protocol will be applied to a nuclear physics example (fragmentation of hot nuclei). It will be shown how it can give new insight in a particular reaction mechanism.