摘要
本文针对缺失数据提出几种差补方法 ,通过模拟实验 ,考察文些方法的适用性及优缺点。结果表明 ,控制变量的恰当引入有利于提高估算效果。从与真值的拟合角度看 ,均值差补法有优势 ;而从保持样本分布的角度看 ,含有随机过程的差补法效果显著。在使用差补后的“完整”数据集时始终保持客观谨慎的态度是非常重要的。
In this paper,on the basis of introduction of some imputation methods, their advantages,disadvantages and adaptable circumstance are discussed through similar test. The result shows that on the aspect of precise mean imputation methods are better than other methods ,while methods containing random process do better on the term of maintaining the distribution of the original sample.Controlling variables may improve the result of estimation while properly used.
出处
《数理统计与管理》
CSSCI
北大核心
2000年第4期50-54,共5页
Journal of Applied Statistics and Management
基金
国家发文委人文社会科学研究"九五"规划项目资助
