空间抽样的理论方法与应用分析——以厦门岛问卷调查为例  

Survey Site Selection Based on the Spatial Sampling Theory:A Case Study in Xiamen Island

作　　者：高丽玲[1] 李新虎[1] 王翠平[1] 邱全毅[1] 崔胜辉[1] 赵千钧[2] 

机构地区：[1]中国科学院城市环境研究所城市环境与健康重点实验室,厦门361021 [2]中国科学院遥感应用研究所,北京100101

出　　处：《地球信息科学学报》2010年第3期358-364,共7页Journal of Geo-information Science

基　　金：中国科学院重要方向性项目(KZCX2-YW-450);国家自然科学基金项目(40901243)

摘　　要：在社会问卷调查中,面临调查样本点如何布局的问题,传统的空间抽样理论主要针对单目标抽样,而社会调查问卷往往涉及到多个调查目标,并且可能是类别数据,例如涉及到被调查对象的年龄、受教育程度等,因此需要在传统的抽样方法上进行适当的改进。本文从多目标空间变异的思想出发,针对社会多目标抽样调查,利用三明治空间抽样模型软件,以厦门岛住区问卷调查的需求为目标,平衡抽样精度和抽样成本,采用简单随机抽样、空间分层抽样和三明治空间抽样方法进行抽样。通过对比分析,结果表明:(1)先验知识不丰富的情况下,简单随机抽样的结果较合理,当有足够先验知识的情况下,进行空间分层抽样能使抽样的样点更具有代表性,提高抽样的精度。(2)空间抽样样点的分布主要与分层相关,分层等级越高,也就是复杂程度越高的区域分布的样点也就越多。(3)通过综合因素分层抽样和专家分层抽样的对比,分层的合理与否,与抽样结果的合理与否有直接的关联,在抽样过程中对分层的把握非常重要。本研究的目的是为厦门岛住区问卷调查选点提供依据,同时,通过复杂程度来描述各个区域的差异,也为建立多目标空间抽样模型提供了思路。We often run into the problem of how to distribute sample points in the process of social survey.Traditional spatial sampling theory is primarily used in single-target.However,social survey questionnaires sometimes are multi-objective.And some objectives are attribute data,for example,a survey related to people＇s age,education and other social information.Therefore,we should make some improvements based on traditional sampling methods to meet current requirements.Depending on the idea of multi-object and spatial variation,we make a multi-objective survey in this study.In order to meet the need of sampling in the questionnaire survey of Xiamen Island＇s settlement,we use the sandwich spatial sampling model,balancing the cost of sampling and sampling precision.In general,we adopt three methods： simple random sampling,spatial stratified sampling and sandwich spatial sampling.After comparing of those three methods,the results showed that：（1） When prior knowledge is not sufficient,the result by means of simple random sampling is more reasonable.But when there is sufficient prior knowledge,stratified sampling in space can make more representative samples for the sampling,and also can improve the accuracy of sampling.（2） The distribution of sample points of spatial stratified sampling is primarily related to stratification.The higher the stratified level,i.e.the higher the complexity of the regional,the more the number of sample points is.（3） By comparing the results of spatial stratified sampling by integrated factors and by experts,we can see that stratified sampling is directly or indirectly related with the stratifying process,the rationality of the stratifying design affects the rationality of sampling results.So,in the process of spatial stratified sampling,the rationality of stratifying design is very important for the stratified sampling.The objective of this study is to provide a basis for the sampling in the settlement of Xiamen Island,offer a concept of ＂complexity＂ to describe the situ

关 键 词：空间抽样 多目标抽样 社会抽样调查 厦门 

分 类 号：O212.2[理学—概率论与数理统计]