空间抽样模型及应用对比分析——以杭州市老年人口抽样调查为例

Spatial Sampling Model and Application Comparative Analysis—Taking the Sample Survey of the Elderly Population in Hangzhouas an Example

摘要: 日前数字化浪潮席卷全国，人口老龄化程度不断加深，数字鸿沟问题更加尖锐，基于杭州成为数字化优势城市的考虑，本文对杭州市老年人口刷脸服务的使用现状开展抽样调查，希望以刷脸服务为切入点探究数字鸿沟问题。考虑到空间异质性和空间自相关性，利用三明治空间模型进行抽样布点，布点因素有老年人口密度和县级行政区人均GDP，并与简单随机抽样和空间分层抽样进行比较，最终得出以下三点结论：1) 在缺乏先验信息的情况下，使用简单随机抽样得出的结果较为合理。当具有一定量的先验信息时，空间分层抽样可以提高抽样布点的精度。2) 从结果来看，综合因素分层抽样和三明治空间抽样较为相似，影响因素的分层很大程度影响了抽样的布点，重要程度越高的层级得到的样本点越多，得到的调查信息越丰富。3) 人的经验知识等先验信息不同会导致分层不同，可以使样本集中在更加有调查意义的区域，故分层前的先验信息尤为重要。

Abstract: Recently, the digital wave has swept the country, the aging of the population has been deepening, and the digital divide is more acute. Based on the consideration of Hangzhou's digital advantage, this paper conducts a sample survey on the use status of face brushing services for the elderly population in Hangzhou, hoping to explore the digital divide problem from all the starting points of face brushing services. Considering the spatial heterogeneity and spatial autocorrelation, the sandwich spatial model was used to sample the distribution points, and the distribution factors were the density of the elderly population and the per capita GDP of the county-level administrative districts, and compared with simple random sampling and spatial stratified sampling, the following three conclusions were finally obtained: 1) In the absence of prior information, the results obtained by simple random sampling are reasonable. When there is a certain amount of prior information, spatial layered sampling can improve the accuracy of sampling distribution points. 2) From the results, the stratified sampling of comprehensive factors and sandwich space sampling are similar, and the stratification of influencing factors greatly affects the distribution of sampling, and the higher the importance level, the more sample points are obtained, and the richer the survey information obtained. 3) Different prior information such as human experience knowledge will lead to different stratification, which can concentrate the sample in a more investigative area, so the prior information before stratification is particularly important.