摘要
农业经济调查缺失数据是一个很常见而又很容易被忽略的问题。在一般缺失模式下,文章利用多元正态模型下的联合分布法对其进行多重插补,拥有很好的估计检验效果。模拟分析显示,根据该方法多重插补后的汇总估计量跟完整数据的估计量非常接近,只是由于数据缺失造成的误差增加使检验显著性下降。跟成列删除后数据的估计检验结果相比,其估计准确性和检验显著性都更高。
Missing data in agricultural economic survey is a normal and ignorable issue. In a general missingness mode, this paper utilizes joint distribution method under multivariate normal model to conduct multiple imputation, and obtains a good result of estimate and test. Simulation analysis shows that, according to this method, the aggregate estimate after multiple interpolation is very close to the estimate of the complete data; but the increase of the error caused by data missing decreases the significance of inspection; the accuracy and significance of the estimation are higher than the estimated test results of the column-deleted data.
作者
潘传快
韩京芳
熊巍
祁春节
Pan Chuankuai;Han Jingfang;Xiong Wei;Qi Chunjie(School of Economics, Wuhan Textile University, Wuhan 430200, China;College of Economics and Management, Huazhong Agriculture University, Wuhan 430070, China)
出处
《统计与决策》
CSSCI
北大核心
2018年第11期12-17,共6页
Statistics & Decision
基金
国家社会科学基金资助项目(16BJY136)
国家现代农业(柑橘)产业技术体系(MATS)专项经费资助项目(CARS-27-08B)
湖北省教育厅人文社会科学基金资助项目(18Y077)
华中农业大学研究生课程建设项目(2015KJ15)
关键词
农业经济调查
缺失数据
多重插补
模拟分析
agricultural economic survey
missing data
muhiple imputation
simulation analysis
