摘要
在传统空间抽样理论中,地理位置信息被用于选取具有空间代表性的样本。当研究目标涉及多个相关指标时,依赖单一辅助变量的传统空间抽样策略往往无法全面捕捉目标变量的多维特征。文章提出了多变量与规模成比例概率(MPPS)空间平衡抽样方法,该方法将空间信息整合到多目标调查中,综合利用总体单元的地理位置信息与属性信息,抽取具有代表性的多目标调查样本。模拟研究分别从总体空间趋势差异、属性变量相关性差异两个维度展开,以期通过构造不同角度、不同水平的目标总体与属性变量来分析新方法的抽样估计效果;并利用全国第四次经济普查中天津市第三产业的企业数量对就业人数进行实证分析。模拟研究和实证结果表明,在多种空间趋势中,新方法均具有不同程度的改进效果,总体空间自相关性越强,属性相关性越强,新方法的改进效率越高;当多个目标的空间分布趋势不一致时,新方法仍有稳健的估计性质。
In the traditional spatial sampling theory,geographic location information is used to select spatially representative samples.However,when the research objectives involve multiple related indicators,the traditional spatial sampling strategies that rely on a single auxiliary variable often fail to adequately capture the multi-dimensional properties of the target variables.This paper proposes the Multivariate Probability Proportional to Size(MPPS)spatial balanced sampling method.The method combines spatial information with multi-objective surveys,and utilizes geographic location information and attribute information of the population units to draw representative samples.The simulation study is conducted from two dimensions:differences in overall spatial trends and differences in attribute variable correlations,aiming to examine the sampling estimation effectiveness of the new method by constructing target populations and attribute variables from various perspectives and levels.Finally,an empirical analysis is performed by using the number of enterprises in the tertiary industry in Tianjin from the Fourth National Economic Census to estimate employment figures.Simulation studies and empirical results show that the new methods have different degrees of improvement under different spatial trends.The stronger the spatial auto-correlation and the stronger the attribute correlation,the higher the improvement efficiency of the new method.When the spatial distribution trends of multiple objectives are inconsistent,the new methods still have robust estimation properties.
作者
杨雪
廖颜玉
张颂
刘丽
Yang Xue;Liao Yanyu;Zhang Song;Liu Li(School of Statistics,Tianjin University of Finance and Economics,Tianjin 300222,China)
出处
《统计与决策》
北大核心
2025年第14期41-46,共6页
Statistics & Decision
基金
国家社会科学基金一般项目(23BTJ058)。
关键词
多目标调查
空间平衡抽样
空间自相关
辅助变量
multi-objective survey
spatial balanced sampling
spatial auto-correlation
auxiliary variables
