大数据背景下多重抽样框方法探讨

Discussion on Method of Multiple Sampling Frame Under Big Data Background

当前所获取的大数据并非都是总体数据,通常未能完全覆盖总体,因其多源异构的特性,致使传统的数据分析方法受阻。文章将抽样调查方法引入到大数据中,对大数据背景下应用多重抽样框的必要性进行剖析,并主要针对大数据中数据多源异构的难点,将每个来源数据作为一个抽样框进行处理,提出了大数据中多重抽样框的构建。进而根据大数据的数据特征进行分类,针对不同情况确定是否需要进行分阶段抽样设计,并提出运用SF估计量对基于多重抽样框的总体进行估计,此估计量较为符合大数据中多重抽样估计的需求,并能对总体有较好的估计。

The big data currently acquired are not all general data,and usually fail to cover the whole,for their multi-source and heterogeneous characteristics cause the traditional data analysis method to be blocked.This paper introduces the sampling survey method into big data and analyzes the necessity of applying multiple sampling fames in the background of big data,and also emphatically aims at the difficulty of multi-source heterogeneity in big data to treat each source data as a sampling frame,proposing the construction of multiple sampling frames in big data.Then the paper classifies them according to the characteristics of big data,and determines whether a phased sampling design is needed according to different situations,proposing to use SF estimator to make an estimation on the whole based on multiple sampling frames.This estimator is in line with the demand of multiple sampling estimation in big data,and the overall good estimation can be achieved.