1: \begin{abstract}

2: The transition density of a diffusion process {does not admit an

3: explicit expression} in general, which prevents the full maximum

4: likelihood estimation (MLE) based on discretely observed sample paths.

5: A{\"{\i}}t-Sahalia [\textit{J. Finance} \textbf{54} (1999) 1361--1395;

6: \textit{Econometrica} \textbf{70} (2002) 223--262] proposed asymptotic

7: expansions to the transition densities of diffusion processes,

8: which lead to an approximate maximum likelihood estimation (AMLE) for

9: parameters. Built on A{\"{\i}}t-Sahalia's [\textit{Econometrica}

10: \textbf{70} (2002) 223--262; \textit{Ann. Statist.} \textbf{36}

11: (2008) 906--937] proposal and analysis on the AMLE,

12: % considered the asymptotic behavior of this estimator.

13: %He pointed that (i) for fixed sampling interval, there exist a

14: %sequence about the approximate terms depended on the sample size

15: %such that the difference between the corresponding AMLEs and the

16: %full MLE convergence to zero in probability; (ii) for fixed

17: %approximate term, there exist a sequence about the sampling

18: %intervals depended on the sample size such that the consistency of

19: %the corresponding AMLEs hold.

20: we establish the consistency and convergence rate of the AMLE, which

21: reveal the roles played by the number of terms used in the

22: asymptotic density expansions and the sampling interval between

23: successive observations. We find conditions under which the AMLE

24: has the same asymptotic distribution as that of the full MLE.

25: A first order approximation to the Fisher information matrix is

26: proposed.

27: \end{abstract}