Continuous Empirical Characteristic Function Estimation of Mixtures of Normal Parameters
提出一种基于连续经验特征函数的高效方法，用于估计离散正态混合模型的参数，通过迭代闭式距离函数提高效率，蒙特卡洛模拟显示其有限样本性质良好，尤其在最大似然估计不收敛时优于离散方法。
This article develops an efficient method for estimating the discrete mixtures of normal family based on the continuous empirical characteristic function (CECF). An iterated estimation procedure based on the closed form objective distance function is proposed to improve the estimation efficiency. The results from the Monte Carlo simulation reveal that the CECF estimator produces good finite sample properties. In particular, it outperforms the discrete type of methods when the maximum likelihood estimation fails to converge. An empirical example is provided for illustrative purposes.