带有离散缺失协变量的Logistic模型的SAEM算法研究

Research on SAEM Algorithm for Logistic Model with Discrete Missing Covariates

给出了带有离散缺失协变量的Logistic模型的一种参数估计方法,首先将离散数据伪连续化,再利用SAEM算法对其进行参数估计,由此提出了一种新的PC(Pseudo Continuous)-SAEM算法.将因变量的回判准确率作为一种新的判别标准,与参数的标准误差、绝对误差和均方误差一起衡量PC-SAEM算法的性能,并将该算法与半参数方法进行对比研究.结果表明,当缺失率小于20%时,PC-SAEM算法的处理性能很好,相比于半参数方法,PC-SAEM算法在运行时间,回判准确率等方面都更加具有优势.最后将该算法应用于实证研究中.

This paper presents a parameter estimation method of logistic model with dis-crete missing covariate data.Firstly,the discrete data is pseudo continuous,and then the SAEM algorithm is used for parameter estimation.Therefore,a new PC(Pseudo Continuous)-SAEM algorithm is proposed.The back judgment accuracy of dependent variable is used as a new criterion to measure the performance of PC-SAEM algorithm together with the standard error,absolute error and mean square error of parameters,and the method is compared with semiparametric method.The results show that when the missing rate is less than 20%,the processing performance of PC-SAEM algorithm is very good.Compared with semiparametric method,PC-SAEM algorithm has more advantages in running time and back judgment accuracy.Finally,the algorithm is applied to empirical research.