摘要
缺失数据在公共突发事件的应急统计数据分析中是一个非常普遍的问题,针对公共突发事件应急统计数据的纵向数据集,提出用一种得分匹配法来进行缺失值的借补处理,并将其与另外三种缺失值处理方法进行比较,即构造各种不同缺失率的随机缺失数据集,分别运用得分匹配法、LVCF借补法、无条件均值抽取法和多重借补法四种不同的缺失值处理方法对每一种缺失率的数据集缺失值进行处理。统计分析结果表明,少数缺失值发生时,LVCF法简单而有效;随着缺失率的增加,均值抽取法和多重借补法处理效果更稳定;得分匹配法借补缺失值考虑了变量之间的相关性,最大程度地利用了数据集包含的信息,同时考虑了含缺失值变量的实际变异程度,因此取得了最好的借补效果。
Data with missing values is common in the emergency statistic under the public emergency. This paper proposes a method named score- match to deal with the longitudinal data sets with different rates of missing values, which has been compared with three other methods such as LVCF, unconditional - mean - distill and multiple imputation. The results of statistic analysis suggest that LVCF is simple and effective when few missing values happen, and with the rates of missing values increasing unconditional - mean - distill and multiple imputation are better. Overall, score- match gets the best result because it takes the relevance and variation between the variables into consideration.
出处
《统计与信息论坛》
CSSCI
2009年第11期3-8,共6页
Journal of Statistics and Information
基金
江苏省软科学研究计划项目<江苏省公共突发事件应急统计的方法和运行机制研究>(BR2008070)
关键词
公共突发事件
应急统计
纵向缺失数据
得分匹配法
public emergency
emergency statistic
longitudinal data with missing data
score- match method
