“…A variety of strategies have been proposed to estimate θ, including simulation-based approaches such as Indirect Inference (Gouriéroux, Monfort and Renault, 1993) or the Efficient Method of Moments (Gallant and Tauchen, 1996), Generalized Method of Moments approaches (Carrasco, Chernov, Florens and Ghysels, 2002;Duffie and Glynn, 2001), nonparametric (Aït-Sahalia, 1996a,b;Bandi and Phillips, 2003;Stanton, 1997) and Bayesian strategies (Eraker, 2001;Jones, 1999), among many others. A few others have been advanced to approximate the unknown transition density, and hence to allow efficient (albeit approximate) maximum likelihood estimation: among these, numerically solving the Fokker-Planck-Kolmogorov partial differential equation (Lo, 1988), closed-form analytic approximation based on Hermite polynomials expansion (Aït-Sahalia, 1999, 2002, and the simulation-based, Monte Carlo integration strategy suggested by Pedersen (1995) and Brandt and Santa-Clara (2002), and further explored by Durham and Gallant (2002).…”