一类双线性系统小样本极大似然估计法

A MAXIMUM LIKELIHOOD ESTIMATION METHOD OF A BILINEAR SYSTEM UNDER SMALL SAMPLES

针对一类双线性系统模型,提出了小样本条件下的极大似然参数估计法。该方法基于预测拟合优势的概念-即对正态分布的最佳拟合是学生公布,而在小样本下其拟合的优势是更为显著的。我们按照此概念构造了具有学生分布的似然函数,用DFP变尺度算法求其极值,仿真表明了该法的有效性。

This paper describes a method of maximum likelihood estimation for a class of bilinear systems under small samples. The method is based on the concept of the superiorty of prediction; i. e. the best fit for the normal distribution is the distribution of students and the superiority of fit even mroe conspicuous under small Samples. According to the concept we construct the likelihood function with student distribution and make it approach the maximum using DFP variable-scale algorithm. The concrete algorthms of the method are given. Good results are shown in the simulation examples.