Recent advances in multi-electrodes array acquisition has made it possible to record the activity of up to several hundreds of neurons at the same time and to register their collective activity (spike trains). This opens up new perspectives in understanding how a neuronal network encodes the response to a stimulus, and what a spike train tells up about the network structure and nonlinear dynamics. For this, one has to develop statistical models properly handling the spatio-temporal aspects of spike trains, including memory effects. In this talk, I will review several such statistical models, including Maximum Entropy Models, Generalized Linear Model or neuromimetic models, and their application for the analysis of retina data.