“…If, further, this variable has a density which depends smoothly on the parameter of interest, η, the Fisher information at stage n has the form I n, (η) = nI (η), where I (η) > 0 is the Fisher information of the model based on the observation of the single variable X , we have the LAN property with rate √ n, the asymptotically efficient estimators η n are those for which √ n( η n − η) converges in law to the normal distribution N(0, I (η) −1 ) and the MLE solves the problem (see, e.g., [5][6][7]). In this setting, a variety of other methods have been proposed in the literature: using the empirical characteristic function as an estimating equation (see, e.g., [8,10,18,19] and Chapter 4 in [23]), maximum likelihood by Fourier inversion of the characteristic function (see [9]), a regression based on the explicit form of the characteristic function (see [14]) or other numerical approximations (see [16,17]). Some of these methods were compared in [3].…”