Edge Localized Modes (ELMs) play a crucial role in present-day tokamak operation. Indeed, the Type-I ELMy H-mode has been chosen as the standard operation scenario for ITER. Thus, understanding and controlling ELMs are important issues for future machine design. In this paper, the dynamic characteristics of Type-I ELM time-series from the JET tokamak have been investigated. The dynamic analysis has been focused on the detection of nonlinear structure in Da time series. The method of surrogate data has been applied to evaluate the statistical significance of the null hypothesis of static nonlinear distortion of an underlying Gaussian linear process. Several nonlinear statistics have been evaluated, such us the time delayed mutual information, the correlation dimension and the maximal Lyapunov exponent. The obtained results allow us to reject the null hypothesis, giving evidence of underlying nonlinear dynamics. Moreover, no evidence of low-dimensional chaos has been found; indeed, the analysed time series are better characterized by the power law sensitivity to initial conditions which can suggest noise contaminated quasi-periodic motion or the "edge of chaos" at the border between chaotic and non-chaotic dynamics.