规模以下工业抽样调查中代表性样本的一种探索设计:平衡抽样设计

Exploring Representative Sample in Below-scale Industry Survey:Balanced Sampling

规下工业抽样调查是社会经济统计调查的重要组成部分,为国民经济核算提供基础数据,而样本代表性直接决定统计推断结果。对企业目录库抽取平衡样本,能够使得样本结构与总体结构相似。平衡样本是指满足如下条件的样本:辅助变量的汉森赫维茨估计等于总体总量真值。平衡抽样设计需要包含丰富辅助信息的完善抽样框,政府统计数据能够为此提供足够的支撑。基于2009年工业企业数据库的实证分析表明,平衡抽样设计对总体总量的估计相对误差很小,特别是估计的均值与总体真值非常接近,近似无偏;与简单随机抽样比较,平衡抽样设计更加有效。

Below-scale industry survey are an important part of the socio-economic survey system, and provide basic data for national account. The representative samples directly determine the quality of statistical data. Business Directory use balanced sampling that the sample structure consistent with that of population. Samples such that the Horvitz-Thompson estimators of the totals for auxiliary variables exactly match the known population totals are called balanced samples. Balanced sampling need perfect sampling frame that is full of Auxiliary information from administrative data. The results of empirical analysis based on 2009 below-scale Business Directory survey shows that small sampling error of target estimator of balanced sampling, which overall mean close to the true value of population particularly, and eventually indicate approximately unbiased estimator. The balanced sampling design is more effective compare to simple sampling.